Impaired CD4 and CD8 Effector Function and Decreased Memory T Cell Populations in ICOS-Deficient Patients

Members of the CD28 family play an important role in the regulation of T cell immune responses (1, 2). Expression of these molecules and their ligands is tightly regulated to deliver either costimulatory or inhibitory signals (2, 3, 4, 5), and their uncoordinated regulation leads to the development of immunological disorders (6, 7, 8).

ICOS (CD278) is a costimulatory member of the CD28 family, and its expression is induced in CD4 T cells upon activation (9, 10, 11). The ICOS signal is induced by interaction with its partner, the ICOS ligand (ICOS-L²; CD275), a molecule highly expressed on B cells and dendritic cells and weakly on T cells and nonlymphoid cells (1, 12). Signaling through ICOS enhances T cell proliferation, secretion of cytokines, and up-regulation of cell surface molecules (11, 13, 14).

Previous research has showed that the ICOS-ICOS-L interaction is important for productive T-B cell coactivation, CD40-mediated Ig class switch recombination, and development of Th2 immune responses (1, 15, 16, 17). Induction of the Th1 cytokine IFN-γ is relatively unaffected and in some studies augmented; other studies have documented the importance of ICOS in Th1 responses (15, 16, 18, 19, 20, 21). Accumulating evidence indicates that ICOS also regulates the generation of Th17 cells, differentiation of FoxP3⁺ regulatory T cells (Tregs), and homeostatic survival of invariant NKT (iNKT) cells (22, 23, 24).

Although earlier investigations of ICOS-null mice revealed normal numbers of naive/memory T cells and normal primary clonal expansion and survival of memory T cells, more recent investigation has demonstrated lower numbers of effector memory T cells (TEMs) in ICOS^−/− mice in the steady state (23). Seemingly contradictory results have been reported on the requirement of ICOS for T cell differentiation and function.

Most studies have depicted ICOS as a costimulator. Indeed, the blockade of the ICOS-ICOS-L interaction abrogates the development of murine models of autoimmune diseases, as follows: rheumatoid arthritis (RA), inflammatory bowel disease (IBD), myasthenia gravis, type I diabetes mellitus, experimental myositis, autoimmune carditis, and graft-vs-host disease (25, 26, 27, 28, 29, 30).

Previously, human ICOS deficiency has been reported in nine patients from four families (31, 32, 33). Importantly, the same homologous genetic deletion of exons 2 and 3 was identified in all patients, indicating a founder effect in all four families. Analysis of these patients revealed reduced numbers of memory B cells and pan-hypoimmunoglobulinemia, but no impairment in the secretion of TNF-α, IFN-γ, IL-2, IL-4, IL-10, or IL-13. Normal surface expression of CD69, CD40L (CD154), CD25, and OX40 (CD134) was observed on their T cells following stimulation (31). A later study provided evidence of defects in IL-10 and IL-17 production (33); however, no major impairment of T cell function was demonstrated. Autoimmunity, manifested as autoantibody-mediated neutropenia, was observed in only one patient (33). Although there have been reports on the effects of ICOS on CD8 responses in mice (34, 35), impact of ICOS on CD8 T cells is not yet completely understood.

In this study, we describe the case of two siblings having ICOS deficiency with a novel mutation in the ICOS gene. Although both patients displayed varing degrees of immunodeficiency, only the sister showed a wide range of autoimmune diseases, including RA, IBD, interstitial pneumonitis (IP), and psoriasis.

In this study, we focused on the T cell immune function of these ICOS-deficient patients. Detailed analysis demonstrated a reduction in memory T cells and a major subtype of Tregs; impaired polarization into Th1, Th2, and Th17; and defective induction of CTLA-4 molecules and other surface inhibitory receptors.

We further assessed activation-induced T cell proliferation and apoptosis, induction of costimulatory receptor molecules, and expression of master regulators for effector T cell subsets, and explored the mechanisms of T cell defect and autoimmunity in these patients using quantitative mRNA analysis.

Patients were diagnosed with common variable immunodeficiency, according to the European Society for Immunodeficiencies criteria (www. esid.org). Twelve healthy volunteers (6 male, 6 female) aged between 26 and 48 years were recruited. The study was approved by the institutional ethical committee of Tokyo Medical and Dental University, and written informed consent was obtained from the patients, the elder sister and mother of the patients, and healthy controls.

Genomic DNA was extracted from peripheral blood using a DNA blood mini kit (Qiagen), according to the manufacturer’s instructions. The coding sequences of the five exons and the adjacent intron-exon boundaries of the ICOS gene were amplified with specific primers (sequences are available upon request) from genomic DNA on the basis of ICOS sequences obtained from GenBank database (accession numbers AC103880, AC009965, and AB023135). All PCR products were sequenced using BigDye terminator v3.1 and an ABI Prism 3130 Genetic Analyzer (Applied Biosystems); the sequence data were then analyzed using DNASIS software (Hitachi Software). Total RNA was isolated from stimulated PBMCs using an RNeasy mini kit (Qiagen) and reverse transcribed into cDNA using a Superscript III first-strand synthesis system for RT-PCR (Invitrogen). PCR products were separated by agarose gel electrophoresis.

We used the following FITC-, PE-, PE Texas Red (ECD)-, or PE cyanin 5.1 (PC5)-conjugated Abs: FoxP3 (236A/E7) from Abcam; IgG1 isotype controls, CD3 FITC, CD3 PE (SK7), CD4 PE (SK3), CD8 FITC, CD8 PE (SK1), CD25 FITC (M-A251), IL-4 PE (3010.211), IFN-γ PE (25723.11), 4-1BB PE (4B4-1), OX40 PE (ACT35), and IL-10 PE (JES3-10F1) from BD Pharmingen; CD3 FITC (UCHT1), CD4 PC5 (13B8.2), CD8 ECD (SFCI21Thy2D3), CD8 FITC (B9.11), CD19 ECD, CD19 PE (J4.119), CD20 FITC (B9E9), CD25 PC5 (B1.49.1), CD28 FITC, CD28 purified (CD28.2), CD45RA FITC (ALB11 and 2H4), CD45RO PE, ECD (UCHL1), CD62L ECD (DREG56), CD69 PE (TP1.55.3), CTLA4 PE (BNI3), streptavidin FITC, streptavidin PC5, and TCR Vβ Repertoire kit from Beckman Coulter (CA); CD27 PE (M-T271) and IgD from DakoCytomation; CD45RO PE (UCHL1), IL-17 FITC (eBio64DEC17), B and T lymphocyte attenuator (BTLA) PE (MIH26), programmed death-1 (PD-1) FITC (MIH4), ICOS-L, ICOSL biotin (MIH12), and ICOS FITC (ISA-3) from eBioscience; receptor activator of NF-κB (RANK) PE (9A725) from Imgenex; CD25 PE (4E3) from Miltenyi Biotec; Alexa Fluor 488 goat anti-mouse IgG Ab from Molecular Probes; CCR7 FITC (150503) from eBioscience; and CD3 purified (OKT3) from Janssen Pharmaceutical.

PBMCs were isolated from heparinized blood using Lymphoprep (Axis-Shield), as described previously (36). CD4 T cells were negatively selected from the PBMCs using a StemSep device (StemCell Technologies). Thus, the purity of the collected CD4 T cell population was generally >95%. CD8 T cells were prepared with the same technique yielding >90% pure CD8 T cell population. Separated cells were resuspended in RPMI 1640 (WAKO) supplemented with 10% heat-inactivated FBS (Gemini Biological Products), and incubated at 10⁶ cells/ml in 24-well plates (Greiner Bioscience) with or without stimulants. For stimulation, we used anti-CD28 mAb (at 1 μg/ml) with plate-bound anti-CD3 mAb or 50 ng/ml PMA (Sigma-Aldrich) plus 1 μg/ml ionomycin (Sigma-Aldrich). The cells were incubated in the medium at 37°C in 5% CO₂ for the indicated time periods. IL-2 (Lymphotec) was used at 700 IU/ml with plate-bound anti-CD3 when assessing ICOS expression.

PBMCs, CD4 T cells, or CD8 T cells were stained with the indicated Abs and were analyzed using a FACSCalibur flow cytometer and CellQuest software (BD Biosciences) or an EPICS XL flow cytometer and EXPO32 software (Beckman Coulter), as described previously (37). For intracellular cytokine detection, PBMCs were stimulated with PMA and ionomycin in the presence of GolgiPlug (BD Pharmingen) or brefeldin A (eBioscience) for 5–8 h at 37°C in 5% CO₂. After stimulation, the cells were fixed and permeabilized using a Cytofix/Cytoperm Plus fixation/permeabilization kit (BD Pharmingen). The same permeabilization technique was used to detect CTLA-4 expression. A CellTrace CFSE cell proliferation kit (Molecular Probes) was used for the CFSE assay, and an annexin V FITC/7-AAD kit (Beckman Coulter) for the apoptosis assay.

Negatively selected CD4 T cells or CD8 T cells were incubated with or without stimulants (plate-bound anti-CD3 mAb and anti-CD28 mAb or 50 ng/ml PMA, and 200 nM ionomycin). The supernatants were collected after 24 h and analyzed using ELISA for IL-17, IL-12p40, IL-22, and TGF-β1 (R&D Systems); IL-21 (eBioscience); and human Th1/Th2 cytokines (IFN-γ, IL-2, IL-4, IL-5, IL-6, IL-10, TNF-α, and TNF-β) using a FlowCytomix kit (Bender MedSystems), according to the manufacturer’s instructions. All assays were performed in duplicate.

Total RNA was extracted using an RNeasy mini kit with DNase (Qiagen) and reverse transcribed using random hexamer primers and Superscript III reverse transcriptase (Invitrogen). Real-time quantitative PCR was performed using a 7300 Real-Time PCR system (Applied Biosystems) using an assay-on-demand Taqman probe and primers (Hs00174383 for IL17A, Hs00243522 for RANKL, Hs00203958 for FOXP3, Hs00226053 for RNF128, Hs00909784 for CBLB, Hs00395208 for ITCH, Hs00172872 for EOMES, Hs00193519 for MAF, Hs00231122 for GATA3, Hs00894392 for TBX21, Hs01076112 for RORC, Hs99999901 for 18S, and Hs99999905 for GAPDH), according to the manufacturer’s instructions. Relative expression levels of these genes were normalized according to GAPDH or 18S rRNA expression, using a standard curve method as described by the manufacturer. All samples and standards were tested in duplicate.

Oligonucleotide microarray assay was conducted with total RNA extracted from a total of 1–3 × 10⁶ CD4 T cells stimulated in anti-CD3-coated plates in the presence of anti-CD28 mAb or from unstimulated CD4 T cells, as described previously (38). Data analysis, selection of significant signals, and comparison of the data from multiple samples were conducted, as previously described (38). The results have been deposited in the Gene Expression Omnibus at http://www.ncbi.nlm.nih.gov/geo/(accession number GSE12875).

The sister, hereafter designated patient 1, was born in 1967. In her infancy, she had episodes of prolonged viral infection. In 2001, when she developed a pulmonary abscess following appendectomy, she was diagnosed with common variable immunodeficiency according to the European Society for Immunodeficiencies criteria (at the age of 34), and i.v. Ig treatment was started to maintain the trough IgG level of >4 g/L. In the following years, she developed psoriasis-like cutaneous lesions and arthritis in multiple joints, including bilateral shoulder, wrist, knee, metacarpophalangeal, proximal interphalangeal, and metatarsophalangeal joints. RA was diagnosed on the basis of the findings of proliferative synovitis of multiple finger and toe joints with erosive changes on x-ray examination. Psoriatic arthritis was ruled out based on the joints affected and the x-ray findings. In 2003, she developed abdominal colic, diarrhea, and IP, and had a constantly elevated serum CRP level. Diagnosis of IBD was made upon biopsy of the colon, and both IBD and IP were controlled by prednisolone. She was referred to our hospital in 2006. Methotrexate at 8 mg/week significantly improved not only the articular signs and symptoms of RA, but also the psoriatic skin changes and IBD. The dose of prednisolone was successfully tapered from 15 to 8 mg/day. Since then, she has been on regular Ig supplementation every 2 wk.

The pedigree of the patient is shown in Fig. 1 A. The patient had two siblings: her sister was healthy with no immunological abnormalities, whereas her younger brother (hereafter designated patient 2) developed occasional skin abscesses and mild psoriasis-like cutaneous lesions, and had slightly low levels of IgG (611 mg/dL) when examined at the age of 35. The serum IgG level stays at the same level to date; and he is not yet on Ig supplementation.

A summary of the immunological data of patients 1 and 2, the elder sister, and ICOS deficiency patients reported to date (33) is given in Table I. Patient 1 had a slightly reduced B cell count, whereas patient 2 had a normal B cell count. In both siblings, however, CD27⁺IgD⁻-switched memory B cells were virtually absent in the peripheral blood samples (Table I). The serum samples contained no detectable specific IgG Abs against measles, mumps, or rubella viruses despite a previous record of vaccine inoculation (data not shown). The immunological parameters of patient 2 are unique in that he showed elevated serum IgM (456 mg/dL). The T cells of the patients displayed abundant expression of CD69 and HLA-DR when stimulated via TCRs in the presence of exogenous IL-2 (data not shown), but lacked surface ICOS expression (Fig. 1 B). Activated T cells from the mother (I.2) and elder sister (II.1) displayed normal ICOS induction.

Sequencing of the ICOS gene revealed the homozygous deletion of T at codon 285, which caused a frameshift in the coding region of ICOS and introduced a premature stop codon at aa 121 (F95fsX121) in the patients (Fig. 1, C and D).

Sequencing analysis of the ICOS gene in the elder sister and mother demonstrated a heterozygous mutation (Fig. 1,C). RT-PCR of ICOS mRNA with specific primers amplifying the entire coding region of the ICOS gene (1–597) demonstrated the presence of an ICOS transcript, suggesting the absence of nonsense-mediated RNA decay (Fig. 1 E).

A previous report on human ICOS-deficient patients showed a normal distribution of naive, memory, and effector T cells (31, 33, 39). However, as seen in the representative FACS plots in Fig. 2 A, we observed a substantial reduction in CD4⁺CD45RO⁺ memory cells in the patients compared with age- and gender-matched controls (12.1 and 6.6% for patients 1 and 2, respectively, and 24.5 and 28.9% for controls 1 and 2, respectively). This reduction was seemingly counterbalanced by an increased frequency of naive T cells.

Gated CD4 memory T cells from PBMCs were further analyzed for CCR7 and CD62L expression to define CCR7⁺CD62L⁺CD45RO⁺ central memory T cells (TCMs) and CCR7⁻CD62L⁻CD45RO⁺ TEMs (40). The analysis showed that compared with controls, the patients had 2- to 5-fold fewer TCMs. The reduction in TEMs was more pronounced, with more than 6-fold fewer TEMs in the patients (Fig. 2, B and C).

A decrease in memory T cells was also observed in CD8 T cells. We observed a reduction in both TCMs and TEMs in patients compared with control subjects (n = 5) (Fig. 2, D–F).

Most Tregs express ICOS, and ICOS^high+ Tregs preferentially produce IL-10 (41). In addition, recent studies have demonstrated the importance of ICOS in proliferation and maintenance of the pool size of FoxP3⁺ Tregs (41). We therefore investigated the frequency of Treg cells in the two patients by staining their PBMCs for CD4, CD25, and intracellular FoxP3. Contrary to our predictions, the patients had a normal proportion of CD4⁺CD25⁺FoxP3⁺ Tregs (Fig. 3,A). However, we noted that the expression level of FoxP3, as reflected by the mean fluorescence intensity, was diminished in both patients (Fig. 3, B and C). To ascertain the low FoxP3 expression obtained in the FACS analysis, we evaluated the level of FoxP3 mRNA in a real-time PCR assay. This showed a marked reduction in FoxP3 expression in patient 2 and a slight decrease in patient 1 compared with the normal subjects (n = 7) (Fig. 3 D).

Recent studies have shown that human CD4⁺CD25⁺FoxP3⁺ Tregs comprise two subsets, as follows: IL-10-producing ICOS⁺CD45RO⁺CTLA-4⁺ Tregs and TGF-β-producing ICOS⁻CD45RA⁺CTLA-4^dull+ Tregs (42). This prompted us to examine CTLA-4 and CD45RO expression in the Tregs of ICOS-deficient patients. Fig. 3, E and F, demonstrates that most CD4⁺CD25⁺ Tregs in these patients were of the CTLA-4^dull+ or CTLA-4⁻ subpopulation and expressed CD45RA (data not shown), indicating that the CTLA-4⁺ subset of Tregs that potentially produces IL-10 was severely decreased.

ICOS-null mice showed defective CD40-mediated Ig class switching because of lack of effective CD40L (CD154) up-regulation (15, 16). In contrast, induction of CD40L was normal in the previously reported cases of human ICOS deficiency (33). Up-regulation of 4-1BB (CD137), BTLA (CD272), and CTLA-4 (CD152) was normal in ICOS knockout mice (23), as was that of OX40 (CD134) and CTLA-4 in patients with ICOS deficiency in a previous study (31).

We estimated the expression of these costimulatory and inhibitory receptors on ICOS^−/− T cells. PBMCs from controls and patients were stimulated with PMA/ionophore (data not shown) or anti-CD3/anti-CD28, and the cells were examined at the end of incubation for the expression of TNF/TNFR family proteins (TNFRI (CD120a), TNFRII (CD120b), CD40L, OX40, and 4-1BB) and of CD28 family proteins (CTLA-4, BTLA, and PD1 (CD279)).

The analysis revealed that CD40L expression was induced normally in the patients’ CD4 T cells, indicating that the hyper-IgM phenotype observed in patient 2 was not due to defective induction of CD40L.

T cells were fully activated in the patients at the end of CD3/CD28 stimulation, as evidenced by CD69 Ag expression. The levels of OX40, 4-1BB, TNFRI, and TNFRII were normal on the surface of the T cells in the ICOS-deficient patients (Fig. 4, A and B, and data not shown).

Baseline CD28 expression in CD4 T cells was similar to that of healthy subjects (data not shown). In contrast, the frequency of CTLA-4⁺ CD4 T cells after CD3/CD28 costimulation was markedly reduced in the patients (Fig. 4,C). Combined data from seven age-matched controls showed that CTLA-4 was induced in 47.4 ± 4.9% of CD4 T cells. In contrast, induction was observed in 30.1 and 24.6% of CD4 T cells in patients 1 and 2, respectively (Fig. 4, C and D). Moreover, as seen in the representative plot in Fig. 4 C, the expression level of CTLA-4 in the CTLA-4⁺ population was also diminished in the patients.

Induction of BTLA, another inhibitory receptor with similarities to CTLA-4 (5), was then estimated. The average percentage of BTLA⁺ CD4 T cells was slightly lower in the patients (22.0% for patient 1; 21.4% for patient 2) compared with controls (34.0 ± 8.7%, n = 7) (Fig. 4, C and D).

In the patients, the frequency of CD4 T cells bearing PD1, a molecule that plays a critical role in the induction and/or maintenance of T cell tolerance (1), was similar to that in controls (Fig. 4, C and D). The percentages of PD1⁺ CD8 T cells, which function as inhibitory T cells (43), were slightly reduced only in patient 2 (29.0%) compared with controls (49.8 ± 9.0%, n = 7).

We next assessed the production of a panel of cytokines by a FlowCytomix bead-based multiplex assay, ELISA, or both, after incubation of CD4 T cells purified to >95% with costimulation of the TCR-CD3 complex via CD28.

In contrast to previous data obtained in other cases of human ICOS deficiency, the production of IFN-γ (Th1 cytokine) and IL-4 and IL-5 (Th2 cytokines) was significantly reduced (Fig. 5,A) in the patients than in controls (31, 33). Secretion of IL-10 and IL-17 was impaired in the ICOS-deficient patients, in agreement with the previous report (33). To confirm the Th17 defect in the patients, a real-time PCR analysis was used to quantify IL-17A mRNA induction. The results, shown in Fig. 5,C, demonstrate a significant decrease in relative IL-17A mRNA expression in ICOS-deficient T cells. Furthermore, induction of other cytokines, such as IL-6, IL-12 p40, TNF-α, and TNF-β, in CD4 T cells was impaired to various degrees in the patients (Fig. 5 A).

Interestingly, the synthesis of the different cytokines was not equally affected in the absence of ICOS: the production of IL-2 was within ±1 SD of normal values, and the IL-22 response was similar to that in controls.

To determine whether the observed defects in effector T cell function can be reproduced by direct activation of intracellular signaling, we examined the capacity of lymphocytes to produce cytokines after PMA/ionomycin stimulation. To that end, intracellular IFN-γ, IL-4, and IL-17 were monitored in PBMCs stimulated with PMA/Ca ionophore by flow cytometry. Fig. 6 A shows that the CD4 T cells of the patients elicited markedly reduced Th1, Th2, and Th17 cytokine responses.

To corroborate these results, we measured the level of cytokines using a FlowCytomix kit in purified CD4 T cells incubated with PMA and ionomycin for 24 h. A similar trend was noted, as follows: the production of IFN-γ, IL-5, IL-10, TNF-α, and TNF-β was found to be diminished. The capacity of ICOS^−/− CD4 T cells to produce IL-2 and IL-6, however, was not markedly impaired (supplemental Fig. 1).³

Because CD45RO⁺ T cells are the major producers of IFN-γ, IL-4, and IL-17 (44) (Fig. 6,A), we considered the possibility that the impaired cytokine responses were due to the decrease in memory CD4 T cells in the patients. To test this, we stimulated PBMCs with PMA/ionomycin and tested for intracellular IFN-γ, IL-4, or IL-17 in the CD4⁺CD45RO⁺ population. Fig. 6, A and B, shows that memory T cells of the patients produced less IFN-γ than the controls. In the controls, 30% of CD45RO⁺ memory T cells produced IFN-γ, whereas in the patients, only ∼10% of CD45RO⁺ memory T cells did so. The synthesis of IL-4 and IL-17 in the memory T cell fraction was marginally decreased in the patients, and the decline was not as clear as that observed in IFN-γ production (Fig. 6 A).

Importantly, the inability to produce IFN-γ does not seem to be restricted to CD4 T cells, because a marked reduction in the IFN-γ response was also evident in the CD4-negative population. To assess effector function of CD8 T cells, we directly measured intracellular IFN-γ in CD8 T cells and a CD8⁺CD45RO⁺ population upon stimulation with PMA/ionomycin. The results displayed in Fig. 6, C and D, revealed impaired IFN-γ production from CD8 T cells and memory CD8 T cells from the patients. The production of IFN-γ was also significantly reduced in CD8 T cells stimulated through CD3 and CD28 in the patients (Fig. 5 B).

We next investigated the mechanisms underlying the T cell unresponsiveness in the patients. One potential explanation is that their T cells did not proliferate well or were prone to apoptosis, or both, in the absence of ICOS expression. To examine this possibility, the proportion of cells that underwent PMA/ionophore-induced cell death was assessed by annexin V/7-AAD staining. The results showed that in the patients, this proportion was similar to or rather lower than that in controls. The proliferative capacity of ICOS^−/− T cells, as assessed by CFSE staining, showed that their T cells proliferated normally or even more vigorously in response to CD3/CD28 cosignal, with significantly more cells with multiple divisions, relative to controls (supplemental Fig. 2).³

Although less likely, the absence of the ICOS-ICOS-L interaction during CD3/CD28 costimulation of CD4 T cells may have contributed to impaired cytokine production in the patients. To test the possible contribution of the ICOS signal in cytokine production, we stimulated purified CD4 T cells from healthy controls (n = 5) through CD3/CD28 with or without anti-ICOS-L blocking Ab (45), and measured the level of cytokines in the supernatants. Supplemental Fig. 3, A–C,³ shows that the effect of ICOS blocking is negligible in this cytokine production assay.

Another explanation for the defective production of effector cytokines is that there were fundamental flaws in their development into effector T cell subsets. We therefore investigated the expression of master transcription regulators of Th1, Th2, and Th17 lineage commitments by quantitative real-time PCR.

Purified CD4 T cells were stimulated with PMA/ionomycin for 4 h or anti-CD3/CD28 for 24 h, and the mRNA expression level of T-bet (for the Th1 lineage) (46), GATA3 and MAF (for the Th2 lineage) (47, 48), and RORC/ROR-γt (for the Th17 lineage) (49) was quantified using GAPDH expression as a control, and expressed as relative expression (RE) adjusted for the baseline expression level of healthy controls (n = 4), taken as 1.0. We observed reduced PMA/ionophore-driven T-bet induction in ICOS^−/− CD4 T cells in the patients, and defective induction was more pronounced in patient 2 (Fig. 7,A). Compared with controls (RE, 16.2 ± 5.3), CD3/CD28-induced T-bet expression was decreased in patient 1 (RE, 7.2), whereas the reduction was less marked in patient 2 (RE, 12.4) (Fig. 7 B).

GATA-3 induction was detectable after PMA/ionophore stimulation. In the patients, induction of GATA-3 above the baseline level was virtually absent in CD4 T cells (RE, 1.3 for patient 1, and 1.0 for patient 2) (Fig. 7,A). The RE values of MAF in CD4 T cells in response to PMA/ionomycin and TCR/CD28 in controls were 3.3 ± 1.4 and 4.4 ± 1.0, respectively. In contrast, stimulation-induced MAF expression was virtually absent in both patients (Fig. 7, A and B).

We also observed that the expression levels of RORC in ICOS-deficient CD4 T cells stimulated with CD3/CD28 or PMA/ionomycin were diminished more than 2-fold (Fig. 7, A and B).

Despite poor IFN-γ production by CD8 T cells, we did not observe low T-bet induction in purified CD8 T cells when stimulated by PMA/ionomycin or by anti-CD3/anti-CD28 mAb (Fig. 7 C). We then tested the expression of EOMES, a paralog of T-bet and a transcription factor required for CD8 effector function (50, 51). Although induction was negligible, there was a trend toward lower baseline expression of EOMES (0.23 and 0.69 for patients 1 and 2, respectively).

Several lines of evidence indicate that E3 ubiquitin ligases (Grail, Cbl-b, and Itch) play important roles in the induction and maintenance of T cell tolerance (52, 53, 54, 55). T cell stimulation without costimulation leads to up-regulation of these ligases (56). High expression of these E3 ubiquitin ligases is related to the absence of the expression of the effector-specific transcription factors (56, 57). There has been little research on the ligases in human systems. Because anti-CD3 stimulation did not induce appreciable up-regulation of the ligases, we examined the mRNA level of these molecules at baseline and after PMA/ionophore stimulation using a sensitive real-time PCR assay.

As shown in Fig. 8, baseline expression of the E3 ubiquitin ligases, with the exception of Grail, was detected in the controls and patients. The mRNA levels of Cbl-b and AIP4/Itch in the steady state were significantly elevated in patient 2 (2.5 and 4.3) compared with controls (1.0 ± 0.4 and 1.0 ± 0.6, n = 4) and patient 1 (1.0 and 1.3).

PMA/ionophore induced up-regulation of Cbl-b and Itch, but not Grail, in normal subjects, whereas the induction of Cbl-b and Itch mRNA above the baseline level was negligible in patient 2 (Fig. 8).

Of particular note was a paradoxical down-regulation of Itch, a regulator of NF-κB activation, upon PMA/ionophore stimulation, which was reproducibly observed only in patient 1.

Patient 1 had an autoimmune manifestation and immunodeficiency, whereas patient 2 had mild psoriasis-like cutaneous lesions and mild skin infections. We therefore attempted to uncover differences in T cell functions between two patients.

To explore the dissimilarities in their immune functions, we assessed mRNA expression levels in negatively selected, >97% pure CD4 T cells by comprehensive mRNA expression analysis using a GeneChip before and after stimulation through CD3/CD28.

The expression of most mRNAs from CD4 T cells poststimulation showed a remarkably high correlation between patients 1 and 2. However, we identified >100 genes that were differently expressed between the patients, and sought to identify the gene(s) that may explain the phenotypic difference from these genes. Among them, TNFSF11 (RANKL) showed >2-fold higher expression in patient 1 after stimulation. In addition, RANKL expression was >50% higher in patient 1 than in the controls.

Because RANKL was identified as a candidate key molecule involved in the pathogenesis of RA (58, 59), we quantified RANKL mRNA in CD3/CD28-stimulated CD4 T cells by real-time PCR. The result shown in Fig. 8 B demonstrates that compared with healthy controls, RANKL induction was higher in patient 1, but lower in patient 2.

In this study, we describe broad defects in T cell function in two siblings with a novel deficiency of human ICOS. Most of the abnormalities presented in this study have not been reported in humans, and some have not been reported in the murine model of ICOS deficiency.

The marked decline in two T cell subpopulations, memory CD4 T cells and CTLA-4⁺CD45RO⁺ Tregs, can be explained, at least in part, by a recent observation that the ICOS-ICOS-L interaction plays an important role in the expansion and survival of these effector T cells (23, 60).

With regard to CD4 memory T cells, we observed significant reductions in the numbers of both TCMs and TEMs in the steady state, which were not observed in the previously reported cases of human ICOS deficiency (31, 32, 33). A reduction in the number of TEMs, but not of TCMs, was demonstrated in ICOS knockout mice by Burmeister et al. (23). TEMs were decreased up to 4-fold in the steady state; the decrease was more pronounced in older mice. TEMs and TCMs display significant and intermediate ICOS expression, respectively (23). Through detailed research on expansion, differentiation, and survival of effector T cells in the absence of ICOS, they suggested that ICOS controls the pool size of effector T cells. These data suggest that both memory subsets may require ICOS for proliferation and survival in humans. Therefore, a decline in total memory cells may be observed in ICOS-deficient mice over a longer observation period and/or after recurrent infections. Alternatively, it is equally plausible that the ICOS-ICOS-L interaction plays a pivotal role in commitment to memory T cells.

CTLA-4⁺CD45RO⁺ICOS⁺CD4⁺CD25⁺ Tregs, commonly observed in adults, were virtually absent in our ICOS-deficient patients. The reduction was seemingly counterbalanced by an increased number of CTLA-4⁻CD45RO⁻ Tregs. Although the contribution of ICOS to the expansion and maintenance of Tregs as a whole has been previously reported (23, 41), our observation addresses the role of ICOS in the maintenance of an IL-10-producing memory subset of Tregs, but not TGF-β-producing CTLA-4⁻CD45RO⁻ICOS⁻ naive Tregs. Supporting this is the observation that in mice, ICOS⁺ Tregs display a strict propensity to undergo rapid apoptosis in culture unless signaled by ICOS-L (23).

The decrease in the number of CTLA-4⁺ Tregs may be alternatively explained by defective induction of a gene that regulates Treg development. A recent study in mice has demonstrated that ROR-γt controls the development of IL-10-producing Tregs that coexpress ICOS in addition to CCL20 (61). This finding may suggest that the decrease in CTLA-4⁺CD45RO⁺FoxP3⁺ Tregs is a consequence of reduced induction of ROR-γt/RORC, as observed in the present study.

Another notable T cell defect in our patients was the impaired capacity of their T cells to mount Th1, Th2, and Th17 responses. Reduced cytokine production was observed not only when the patients’ CD4 T cells were activated by costimulatory signals, but also when they were stimulated by PMA/ionomycin.

Although the ICOS-ICOS-L interaction was important in vivo in the generation and/or maintenance of effector memory and central memory cells, the absence of an ICOS signal through ICOS-L did not seem to contribute to the T cell effector defects observed in our ex vivo experiments. First, ICOS-L expression was not induced in purified T cells upon CD3/CD28 costimulation (supplemental Fig. 3A).³ In addition, blocking potential ICOS-ICOS-L interaction in the controls did not result in decreased cytokine production or in decreased up-regulation of MAF and RORC (supplemental Fig. 3, B and C).³

Additional experiments indicated that there was an abnormality at the level of transcriptional regulation of Th1, Th2, and Th17 polarization, and decreased induction of the master regulators T-bet, GATA-3, MAF, and RORC in the patients. Previous research on mice has shown that ICOS regulates MAF expression and GATA-3 induction (62), and our present study points to an additional role of ICOS in the complete induction of T-bet and RORC.

One major factor contributing to the poor effector T cell responses in the patients could be the decrease in total memory CD4 T cells. This is particularly likely in the case of IL-4 and IL-17 production, because the memory T cells had only mild defect in producing IL-4 and IL-17. Although the CD4⁺CD45RO⁺ T cells in the patients displayed a significantly reduced ability to produce IFN-γ, the decreased response may be explained by pronounced reduction in TEMs in the patients. To determine whether the memory T cell compartment in our patients is functionally defective or intact on a per cell basis, we would need further analysis of various parameters of the T cell effector functions in naive T cells, TCMs, and TEMs.

Nurieva et al. (56) demonstrated that murine ICOS^−/− CD4 T cells showed defective induction of T-bet, GATA-3, and EOMES in the absence of CD28 costimulation because of up-regulation of E3 ubiquitin ligases: Grail, Cbl-b, and Itch. It is uncertain whether the augmented baseline expression of E3 ubiquitin ligases is relevant to the observed effector T cell dysfunction, because this was confirmed only in patient 2. It is rather unlikely that the different expression of the E3 ubiquitin ligases contributed to the T cell defects in the patients because augmented induction of these ligases was not detected in the patients.

In contrast to the global impairment in cytokine synthesis, IL-2 production was only marginally affected. Supporting this observation, induction of transcription factors for IL-2 (c-Jun/c-Fos) was normal in the patients (data not shown). Similarly, induction of IL-21 and TGF-β was also unaffected in the CD4 T cells of the patients, although their induction was modest under costimulatory conditions (data not shown). It should be noted that production of IL-22, a Th17 cytokine fundamental for the development of psoriasis, showed normal induction, and that both the patients had psoriatic cutaneous lesions (63). Although whether IL-17A and IL-22 are produced by the same Th17 subset is still unclear (64, 65), our data suggest an IL-22-producing CD4 T cell subset is not functionally impaired in the patients.

Previous studies in mice have shown that ICOS is necessary for optimal CD8 T cell responses (34). ICOS can directly stimulate CD8 T cells (35); and ICOS-Ig-treated mice displayed diminished IFN-γ production by CD8 T cells. Our study has demonstrated that CD8 memory cells are reduced in ICOS deficiency, and that CD8 T cells in the absence of ICOS can mount a very low IFN-γ response, for the first time in humans. ICOS is induced on terminally differentiated CD28⁻CD8 effector T cells (11), and thus, may play a role in maintaining the CD8 subset. Therefore, a decrease in the number of IFN-γ-producing CD8 T cells could be ascribed to the reduction in CD45RO⁺ memory CD8 T cells or in CD28⁻CD8 T cells (data not shown) in our ICOS-deficient patients. IFN-γ production is regulated by T-bet and EOMES cooperatively or redundantly in CD8 T cells (50, 51). T-bet induction was normal when stimulated with PMA/ionomycin or anti-CD3/anti-CD28, whereas a baseline expression of EOMES was decreased in CD8 T cells in the patients. Although this may explain the impaired production in part, further research on the CD8 T cells stimulated with various common γ-chain cytokines would be necessary to assess whether the transcriptional regulation of CD8 effector functions is impaired in the absence of ICOS.

In addition to the reduced numbers of effector T cells, which either potentiate or inhibit T cell responses, the present study demonstrates for the first time an aberrant induction of negative costimulatory molecules on activated T cells in ICOS-deficient patients. CTLA-4 and BTLA are induced upon activation and transmit an inhibitory signal to T cells to regulate the balance between T cell activation, tolerance, and immunopathology (3, 4, 5). Costimulatory and coinhibitory molecules are normally induced in the absence of ICOS in mice and humans (23, 31). In our patients, however, induction of CTLA-4 and BTLA following CD3/CD28 signaling was impaired. Although the molecular basis of the defective expression is still not known, this may be ascribed to the decreased memory T cell subset in the patients. At all events, our findings imply that an inhibitory signal to suppress activated T cells could not be appropriately induced in the patients.

Collectively, these data highlight the positive contribution of ICOS to the maintenance of, or commitment to, effector T cells and a subset of Tregs, and the induction of negative costimulatory receptors on activated T cells. The immunodeficiency in our ICOS-deficient patients, although mild, can be understood by the defects in their effector T cell functions as well as in T cell-dependent B cell help, but a reasonable explanation is still required for the development of autoimmunity, RA, IBD, IP, and psoriasis in ICOS-deficient patients.

Most studies have depicted ICOS as a positive costimulator in the immune reaction. For example, research in ICOS-deficient mice suggests that ICOS is critically involved in autoimmune development and allogeneic reactions (21, 25, 26, 27, 28, 29). There are, however, some results indicating that abrogation of the ICOS-ICOS-L interaction aggravates the disease process. For example, in some initial studies on ICOS-null mice, experimental autoimmune encephalomyelitis was unexpectedly exacerbated and allergen-dependent airway sensitivity was augmented (14, 66, 67). What is the role of ICOS in autoimmune development?

Burmeister et al. (23) reported that ICOS supports the expansion and survival of Th1 or Th2 responder cells, Th17 cells, and FoxP3⁺ regulatory effector cells. They hypothesized that the absence of ICOS function in a particular mouse model would result in a phenotype reflecting a deficiency of the dominant effector T cell type. Thus, blockade of the ICOS-ICOS-L interaction could mainly affect Treg subsets and lead to the development of autoimmune disorders.

Our observations in human ICOS deficiency may fit the concept of ICOS as an agonist molecule. In our ICOS-deficient patients, defects in T cell function leading to termination of the activated T cell response may have been dominant.

Another question remains, as we observed a wide range of autoimmune diseases in patient 1, but not in patient 2. Although phenotypic variation in siblings with the same mutation is not uncommon in human genetic disorders, there might have been some contributing factor(s).

First, our analysis showed paradoxical down-regulation of Itch expression after PMA/ionophore stimulation in patient 1. Because Itch induction is important in the control of NF-κB activation (55), an activation signal in the absence of CD28 costimulation may have led to continuous inflammation in patient 1.

Second, although the T cell immune functions and stimulation-induced mRNA expression pattern of CD4 T cells were strikingly similar between the siblings, we found that the T cells of patient 1 showed exaggerated induction of RANKL expression and poorer production of IFN-γ. Previous studies demonstrated that T cells, which contribute to the development of RA and IBD, were characterized by poor IFN-γ production, production of inflammatory cytokines including IL-17A and TNF, and RANKL expression (59). Although IL-17A induction was not increased, the augmented RANKL expression and poor IFN-γ production in T cells may have contributed to the autoimmune disease progression. Characterization of RANKL-expressing IL-17A-negative T cells requires further investigation.

Third, we surmised that a major infectious episode may have upset the subtle balance between effector T cells and Tregs in our patients. In fact, patient 1 developed a series of autoimmune disorders after a severe bacterial infection.

Finally, the reason for the apparent discrepancy in T cell functions between the ICOS-deficient patients presented in this study and the ICOS deficiency described in previous reports (31, 32, 33) is elusive. One possibility that explains the difference in cytokine production in our patients and ICOS deficiency in previous reports could be different stimulation condition (dose of mAb and incubation period). However, it is unlikely because effective cytokine synthesis from the T cells was not observed in our patients even with an increased dose of anti-CD28 mAb and with longer time periods (supplemental Fig. 4).³ This indicates the presence of other intrinsic factor(s).

Another possibility is the difference in the mutation site of the ICOS gene. Our patients harbored a homozygous single-base deletion at codon 285 located in the extracellular domain, whereas other ICOS deficiencies have homozygous deletion in exons 2 and 3 of the ICOS gene (33). In addition to defective ICOS expression, this mutation may result in the expression of a 120-aa ICOS protein that affects immune function, for example, by binding to ICOS-L on B cells, monocytes, or a subset of T cells. Despite extensive investigation, however, we have been unable to demonstrate the expression of a truncated ICOS protein in our patients’ lymphocytes.

In summary, the present study on T cell functions in two novel ICOS-deficient patients has shown that the interaction between ICOS and ICOS-L is critical for the development and maintenance of multiple types of effector cells and Tregs, and that the defects are at least in part due to diminished memory T cells and/or impaired induction of master regulators. Collectively, the results of our study highlight a major role of ICOS as a coordinator of T cell immune responses and T cell maintenance.

We thank Nakaba Ochiai and Shizuko Minegishi for technical assistance, and Drs. Erdyni Tsitsikov and Jun-ichi Yata for critical reading of the manuscript.

The authors have no financial conflict of interest.