IL-22 Binding Protein Promotes the Disease Process in Multiple Sclerosis

This article is featured in In This Issue, p.773

Multiple sclerosis (MS), a chronic inflammatory disease of the CNS, is still treated with relatively unspecific modulators of the immune system (1, 2). Therefore, lack of response and severe adverse events remain clinical challenges. Development of more appropriate treatments requires pharmacological targeting of disease-specific molecular events while sparing normal immune function. Drug targets that meet these criteria can be identified by characterizing the genetic variants and, in extension, the physiological traits that predispose for the disease. Recent genome-wide association studies have revealed a large number of genetic variants that influence MS susceptibility (3–5). These are frequently found near immune-related genes, but the underlying mechanisms are, with few exceptions, unknown (6–10). Defining the causal genes and their impact on MS pathogenesis could potentially lead to safer and more effective treatments.

In the current study, we sought to clarify the role in MS of a risk gene variant localized downstream of IL22RA2, namely, the single nucleotide polymorphism (SNP) rs17066096. This locus has repeatedly been associated with MS susceptibility (3–5) and encodes the IL-22 binding protein (IL-22BP), an antagonist molecule that tightly regulates the actions of IL-22 (11, 12). Signaling via the IL-22R, highly expressed on epithelia at barrier surfaces such as skin, lungs, and gut, leads to tissue protection and regeneration as well as induction of antimicrobial peptides (13, 14).

A relatively physiological way to characterize the effects of IL-22 is to manipulate the levels of IL-22BP in animal disease models. We and others have shown that IL-22BP acts to limit inflammation in models of psoriasis and inflammatory bowel disease, both affecting barrier surfaces (15, 16). In contrast, we have also shown that both rats with a natural genetic variant in this locus that confers lower expression of IL-22BP and mice with full gene deletion have less severe disease in an animal model of MS (17, 18). However, the role of IL-22 in MS is unclear. IL-22 is often described as a pleiotropic cytokine with a high degree of context dependency regarding its downstream effects (14). Inflammation at barrier surfaces and in the CNS may thus be two contexts in which the net effects of IL-22 are opposite.

In this study, we show that an MS genetic risk sequence variant strongly affected IL-22BP production and immune cell activation state in vitro, as evidence for an expression quantitative trait locus (eQTL). We confirm the disease-promoting role of IL-22BP in rat and mouse MS models and propose a mechanism of action whereby IL-22 acts directly on T cells to limit their encephalitogenic potential. Cytokines and their receptors, either soluble or membrane bound, are often considered as therapeutic targets in a variety of inflammatory diseases. Our data on the IL-22 system provides important background for any attempt to modulate this system as a therapeutic strategy in humans.

For experiments in Fig. 1F, buffy coats were collected from healthy blood donors for which rs17066096 genotype was known. PBMCs were isolated using Ficoll-Pacque PLUS (GE Healthcare) gradient centrifugation. Buffy coat cells were frozen in medium containing DMSO, later thawed to isolate CD14⁺ cells using MACS microbeads (Miltenyi Biotec), and then live frozen again. CD14⁺ cell samples from all healthy donors were thawed and plated in parallel to avoid batch effects. For experiments in Fig. 1A–D, blood was collected from healthy volunteers of unknown rs17066096 genotype and immediately processed for in vitro stimulation.

After thawing, CD14⁺ PBMCs were washed, counted, and resuspended at a concentration of 1.5 × 10⁶ cells per milliliter in RPMI 1640 medium supplemented with 2 mM glutamine, penicillin, streptomycin, and 10% FBS. Two hundred microliters of cell suspension was added per well in 48-well plates. Aliquots of each sample were saved to assess purity of CD14⁺ isolation by flow cytometry, which was consistently >95%. Cells were differentiated for 6 d with a change to fresh medium after 3 d. Cells were harvested by removing the medium and processing the cells for RNA isolation. The following reagents were used for in vitro differentiation of CD14⁺ PBMCs: 400 IU/ml GM-CSF (PeproTech), 20 ng/ml IL-4 (R&D Systems), 50 ng/ml M-CSF (R&D Systems), and 0.001–100 nM AM580 (Sigma-Aldrich). To produce M1 and M2 polarized cells, M-CSF–differentiated cells were additionally stimulated 24 h with fresh medium containing 100 ng/ml LPS + 20 ng/ml IFN-γ and 20 ng/ml IL-4 (R&D Systems), respectively, with or without the addition of AM580. Because AM580 is dissolved in DMSO, an equivalent amount of DMSO (the vehicle) was consistently added to wells that were not differentiated with AM580. For activation of GM-CSF/IL-4–differentiated cells with TLR agonists in Fig. 1C, the following reagents were added 24 h before harvesting the cells: 1 μg/ml PAM3CSK4, 1 × 10⁸ cells per milliliter of heat-killed Listeria monocytogenes, 10 μg/ml poly(I:C) high m.w., 10 μg/ml poly(I:C) low m.w., 100 ng/ml LPS from Escherichia coli K12, 1 μg/ml Flagellin from Salmonella typhimurium, 100 ng/ml FSL1, 10 μg/ml Imiquimod, 10 μg/ml ssRNA40, and 5 μM ODN2006 (Invivogen). In Fig. 4B–D, CD4⁺ and CD8⁺ human PBMCs were isolated and plated at 80,000 cells per well in flat-bottom 96-well plates. Two microliters of Dynabeads Human T-activator CD3/CD28 (Life Technologies) was added to each well, according to the protocol supplied by the manufacturer. Cells were cultured 24 h prior to analysis by quantitative PCR (qPCR). Two micrograms per milliliter of recombinant human IL-22 (R&D Systems) was added from the start, and cells were then cultured for 72 h followed by analysis by flow cytometry. For the experiment in Fig. 4E, blood samples were collected from 84 individuals with MS or other neurologic diseases, followed by PBMC isolation, live freezing, and storage in liquid nitrogen. Fifty thousand cells were plated in round-bottom 96-well plates with cell culture medium alone or with the addition of 10 μg/ml LPS. Cells were harvested after 3.5 h and were processed for RNA extraction and qPCR analysis. Fold change in expression was calculated by dividing relative expression after LPS stimulation with relative expression in control wells.

Genotyping of healthy blood donors for the experiment in Fig. 1E and 1F was performed on the ImmunoChip (Illumina) by the SNP&SEQ Technology Platform, Science for Life Laboratory and the National Genomics Infrastructure at Uppsala University. For the experiment in Fig. 4E, total DNA was extracted from blood samples using QIAamp Blood Maxi kit (Qiagen), according to the protocol provided. rs17066096 genotype was assessed using the TaqMan OpenArray Genotyping System (Applied Biosystems).

The Dark Agouti (DA) rat strain was originally obtained from the Zentralinstitut für Versuchstierzucht (Hannover, Germany), and MHC-identical Piebald–Virol–Glaxo (PVG.1av1, referred to as PVG in this article) was obtained from Harlan UK Limited (Blackthorn, U.K.). Colonies have thereafter been established at Karolinska University Hospital (DA/Kini and PVG.1av1/Kini). An inducible knockdown rat strain on the DA background was generated by TaconicArtemis, project name DA/OlaHsd-Tg(H1/tetO-RNAi:IL22ra2)Arte. A DNA construct containing a short hairpin RNA specific for Il22ra2 under the control of an H1 promoter with a tetO site as well as a TetR under the control of the CAGGS promoter was used (Fig. 3A). The DNA solution was microinjected into one cell stage fertilized embryos from DA/OlaHsd rats (Harlan) and transferred to oviducts of pseudopregnant Sprague Dawley females. Integration of the transgene was detected by PCR on genomic DNA isolated from ear biopsies with the primers 1165_48: 5′-CCA TGG AAT TCG AAC GCT GAC GTC-3′ and 1165_46: 5′-TAT GGG CTA TGA ACT AAT GAC CC-3′. For knockdown experiments, tap water was supplemented with 20–500 mg/l doxycycline (D-9891; Sigma-Aldrich) and 1–10% (w/v) sucrose (Sigma-Aldrich) and was administered ad libitum as drinking water. The solution was protected from light and was freshly prepared every second day.

The mouse strains used for this research project, B6N.129S5-Il22ra2tm1Lex/Mmcd (targeting Il22ra2), identification number 032407-UCD, and B6;129S5-Il22^tm1.1Lex/Mmucd (targeting Il22), identification number 036745-UCD, were obtained from the Mutant Mouse Regional Resource Center, a National Center for Research Resources and National Institutes of Health–funded strain repository, and had been donated to the Mutant Mouse Resource and Research Center by Genentech (18). Il22ra2 is a constitutive knockout strain, and Il22 is a conditional knockout strain using the cre/loxP system. Because IL-22 is only expressed by hematopoietic cells, a full knockout is achieved by introducing cre recombinase under the Vav1 promoter. To avoid bias, all mice used for experimental autoimmune encephalomyelitis (EAE) were bred to express Vav1-cre (The Jackson Laboratory).

Female mice and rats between the age of 12 and 20 wk were used for all experiments. Age-matched littermates were used as controls. Animals were bred and maintained in a specific pathogen–free facility at the Karolinska University Hospital (Stockholm, Sweden) in polystyrene cages with aspen wood shavings, free access to water, and standard rodent chow.

Commercial ELISA kits were used to quantify human and rat IL-22BP (Cloud-Clone Corp) and rat IL-22 (R&D Systems), according to the protocol supplied by the manufacturer. Plasma and CSF were collected and stored at −80°C until analysis.

To induce EAE, rats were injected s.c. at the tail base with 4.5 μg of rat myelin oligodendrocyte glycoprotein (MOG) emulsified in Freund’s adjuvant (Sigma-Aldrich). Mice were injected s.c. at the tail base with 50 μg of mouse MOG emulsified in Freund’s adjuvant supplemented with 200 μg nonviable Mycobacterium tuberculosis H37 RA (Difco Laboratories). At immunization and 2 d later, the mice were injected i.p. with 200 ng of pertussis toxin (Sigma-Aldrich). Animals were weighed and observed daily. Clinical signs were assessed as follows: 0, no paralysis; 1, tail paralysis; 2, hind limb paresis 3, hind limb paralysis; 4, hind and front limb paralysis; 5, dead.

Microbeads for OX-52 (T cells), CD45R (B cells), CD11b/c (myeloid cells), and OX-62 (dendritic cells [DCs]) (Miltenyi Biotec) were used to isolate basic cell types from single-cell suspensions of inguinal lymph node, according to the protocol supplied by the manufacturer.

Mice were perfused with PBS containing 500 μl/l Heparin (Leo Pharma). Brains were dissected and processed using the neural tissue dissociation kit (T) (Miltenyi Biotec), according to the protocol provided by the manufacturer. For magnetic separation of CNS cells, the following microbead kits for mice were used: ACSA-1, ACSA-2, O4, CD11b, and CD171 (Miltenyi Biotec). A dedicated CNS cell sample fraction was used to isolate CD11b⁺ microglia by positive selection. To sort astrocytes, another CNS cell sample fraction was incubated with a pool of ACSA-1 and ACSA-2 microbeads followed by positive selection. The flow-through from this step was then passed through an LD column (Miltenyi Biotec) to achieve an astrocyte-depleted fraction, which was split into two fractions for positive selection of O4⁺ oligodendroglia and CD171⁺ neurons. Purity was assessed by qPCR of Gfap, Mog, Itgam (CD11b), and Rbfox3 (NeuN) (Supplemental Fig. 3B).

Oligodendrocyte progenitor cells were purified by magnetic cell sorting. Brains were obtained from postnatal day 2 Wistar rats (Charles River), and the forebrains were removed by crude dissection. Tissue was diced and digested in papain (Worthington) at 37°C for 60 min and then gently dissociated. Dissociated postnatal day 2 brains were incubated with 20 μl A2B5-coated microbeads (Miltenyi Biotec) per 10⁷ cells for 15 min at 4°C. After rinsing with DMEM containing 1× ITS supplement (Invitrogen), up to 10⁸ dissociated cells were resuspended in 3 ml of DMEM/ITS and applied to an LS magnetic separation column placed in a magnetic field. After washing the column with 3 ml of DMEM/ITS, the positively selected A2B5⁺ cells were eluted from the column by removing the column from the magnetic field, passing 5 ml of complete media (see below) through, and collecting the eluted cells in a collection tube. A2B5⁺ cells were transferred to poly-d-lysine–coated 96-well culture plates and seeded at 10,000 cells per well. All cells were cultured at 37°C, 5% CO₂ in DMEM containing penicillin–streptomycin (100 U each), 1× GlutaMAX supplement, 1× N2 supplement (all from Invitrogen), and modified NS21 (19) including 2.5 mg/ml BSA, 2.5 μg/ml catalase, 1 μg/ml glutathione, 2.5 μg/ml superoxide dismutase, 2 μg/ml l-carnitine, 1 μg/ml ethanolamine, 15 μg/ml d-galactose, 1 μg/ml linoleic acid, 1 μg/ml linolenic acid, 0.047 μg/ml Lipoic acid, 1 μg/ml tocopherol, 1 μg/ml tocopherol acetate, and 10 nM hydrocortisone. The following day, the cells were treated with differentiation factors 10 ng/ml T3 (Sigma-Aldrich) and 10 ng/ml CNTF (R&D Systems) with or without 500 ng/ml recombinant rat IL-22 (R&D Systems). Cells were cultured for a further 6 d with feeding every 2–3 d.

Cells were fixed with 4% formaldehyde (Pierce) for 20 min at room temperature. After washing, cells were permeabilized with 0.1% Triton X/PBS for 5 min. Cells were stained with HCS cell mask green (Invitrogen) before blocking with 3% BSA (Sigma-Aldrich) in PBS and staining with Hoechst 1:10,000 (Invitrogen), anti-MBP 1:1000 (Abcam), and goat anti-mouse secondary Ab Alexa Fluor 647 1:1000 (Invitrogen). Cells were imaged on the OPERA QEHS using 10× air objective and analyzed using Columbus software.

Brains of PBS-perfused wild-type mice were aseptically dissected and collected in 50-ml tubes containing 10 ml of ice-cold HBSS, then transferred into 5 ml of Leibovitz’s L15 medium (Thermo Fisher Scientific) containing 500 U of Papain (Worthington) per brain and mechanically minced by pipetting up and down through a 25-ml pipette. The homogenate was incubated in a water bath at 37°C for 15 min. After 5 min of incubation, remaining tissue pieces were further reduced by trituration with 1-ml tips. After 10 min of incubation, tissue was triturated by pipetting with a 200-μl tip prior to addition of 400 U of DNase I (Roche). The enzymatic reaction was stopped by adding 20 ml of HBSS + 10% FBS (v/v). The homogenate was transferred and filtered through a 40-μm cell strainer into a new 50-ml tube. After centrifugation at 200 × g, the pellet was resuspended in 20 ml of 20% Percoll (Sigma-Aldrich) in HBSS (v/v) and centrifuged for 20 min at 200 × g at room temperature with low acceleration and no break. Following density centrifugation, the myelin layer and remaining supernatant were removed using a 25-ml pipette. The pellet was washed in HBSS at 200 g. A pellet of one to three brains containing mixed glia cells was resuspended in 20 ml of prewarmed DMEM/F12 (Thermo Fisher Scientific) including 10% FBS (v/v), 100 U per ml/100 μg per ml penicillin/streptomycin (Sigma-Aldrich), and 20 ng/ml mouse M-CSF (R&D Systems) and seeded into an uncoated T75 cell culture flask. The medium was changed twice a week until the cells reached 90% confluency (12–16 d).

To obtain astrocytes, mixed glial cells were detached by 10 ml of prewarmed Trypsin-EDTA 0.05% (Thermo Fisher Scientific). After 10–15 min of incubation in Trypsin-EDTA at 37°C, cells were collected through a 40-μm cell strainer into a 15-ml Falcon tube, and the reaction was stopped by adding 5 ml of pure FBS. The single-cell suspension containing both microglia and astrocytes was centrifuged at 300 × g for 10 min. The mixed glia culture was depleted of CD11b⁺ cells using magnetic beads (Miltenyi Biotec). Cells in the negative fraction were resuspended at a density of 10⁵ cells per ml in DMEM/F12 including 10% FBS (v/v) and penicillin/streptomycin and seeded at 1 ml per well in an uncoated 24-well plate (Thermo Fisher Scientific). After 24 h, cells were stimulated in prewarmed complete medium containing either 100 ng/ml LPS (Sigma-Aldrich), 20 ng/ml IFN-γ (R&D Systems), 30 ng/ml TNF (R&D Systems), 3 ng/ml IL-1α (R&D Systems), or 1 mM H₂O₂ or medium without FBS (for starvation) alone or in the presence of 10 ng/ml IL-22 (R&D Systems) for 48 h, after which cells were detached with prewarmed Trypsin-EDTA 0.05%; stained with APC Annexin V and propidium iodide (BioLegend), according to the manufacturer’s protocol; and quantified using flow cytometry.

Cells were disrupted in RLT buffer (Qiagen) with 1% 2-ME and homogenized by repeatedly passing the lysate through a needle with an RNase-free syringe. Tissue samples were disrupted and homogenized with RNase-free steel spheres in RLT buffer with 1% 2-ME using a TissueLyzer LT (Qiagen). Total RNA was extracted using the RNeasy mini kit (Qiagen) with the QIAcube (Qiagen), including on-column digestion of DNA using the RNase-free DNase set (Qiagen). The iScript cDNA synthesis kit (Bio-Rad) was used to prepare cDNA. Real-time RT-PCR was performed on a CFX384 Real-Time system machine (Bio-Rad) using IQ SYBR green Supermix (Bio-Rad). Cycling conditions were as follows: 95°C for 3 min, (95°C for 15 s, 60°C for 30 s, 72°C for 30 s) (40 cycles), 95°C for 10 s, followed by melt curve analysis. Data were analyzed manually using the Δ cycle threshold method with HPRT or Hprt as reference. Primer sequences used are as follows: human HPRT: forward (F): 5′-ACCAGTCAACAGGGGACATAA-3′, reverse (R): 5′-CTTCGTGGGGTCCTTTTCACC-3′; human IL22RA2: F: 5′-ACACTTGCAACCATGATGCC-3′, R: 5′-TGAGGCTTCAGAGACTCATGC-3′; human IL22RA1: F: 5′-AGATATGTCACCAAGCCGCC-3′, R: 5′-CCGCTGAGGTCAAAGACAGG-3′; human IL22: F: 5′-CCAGCATGAAGGTGCGGTTGGT-3′, R: 5′-TGCGCCCATCAGCTCCCACT-3′; rat Hprt: F: 5′-CTCATGGACTGATTATGGACA-3′, R: 5′-GCAGGTCAGCAAAGAACTTAT-3′; rat IL22ra2: F: 5′-CAACCACGATGCCTAAGC-3′, R: 5′-CAACCACGATGCCTAAGC-3′; rat Il22ra1: F: 5′-TACACGTGCCGAGTGAAGAC-3′, R: 5′-GCTGGACATTCAGGGAGTTAGG-3′; rat Il22: F: 5′-CAGGAGGTGGTGCCTTTCC-3′, R: 5′-TCTTCTGGATGTTCTGGTCATCA-3′; mouse Hprt: F: 5′-ACAGCCCCAAAATGGTTAAGG-3′, R: 5′-TCTGGGGACGCAGCAACTGAC-3′; mouse Il22ra2: F: 5′-GAAGGTCCGATTTCAGTCCA-3′, R: 5′-TCACCCTCCCGTAATACAGC-3′; mouse Il22ra1: F: 5′-GCTCGCTGCAGCACACTACCA-3′, R: 5′-TCCAGGGTTAGCTGGTGGCCA-3′; and mouse Il22: F: 5′-GCTCAGCTCCTGTCACATCA-3′, R: 5′-CAGTTCCCCAATCGCCTTGA-3′.

Rat inguinal lymph nodes were collected in cell culture medium and passed through a 40-μm strainer to make single-cell suspensions. Cells were incubated with yellow dead cell staining (Invitrogen) and stained with following Abs: FITC CD3 (clone IF4; BD Biosciences), PE IFN-γ (clone DB-1; BD Biosciences), PerCP/Cy5.5 IL-17A (clone eBio17B7; Thermo Fisher Scientific), PE-Cy7 Foxp3 (clone FJK-16s; Thermo Fisher Scientific), and APC CD4 (clone OX-35; BD). Human PBMCs were isolated and used for in vitro experiments, as described, and were stained with the following Abs: PE/Dazzle CD3 (clone OKT3; BioLegend), Alexa Fluor 700 CD4 (clone A161A1; BioLegend), Alexa Fluor 700 CD8a (clone HIT8a; BioLegend), PerCP/Cy5.5 IL-17A (clone BL168; BioLegend), PE IFN-γ (clone B27; BioLegend), APC Foxp3 (clone PCH101; Thermo Fisher Scientific), and V450 Ki-67 (clone B56; BD Biosciences). Foxp3 staining kit (Thermo Fisher Scientific) was used for intracellular staining. Abs used for spinal cord–infiltrating cells in mice were as follows: PE-Cy7 CD45.2 (clone 30-F11; BioLegend), PerCP/Cy5.5 CD11b (clone M1/70; BioLegend), V500 CD3 (clone 500A2; BD Biosciences), FITC CD11c (clone HL3; BD Biosciences), V450 Ly6G (clone 1A8; BD Biosciences), and near-IR dead cell staining dye (Invitrogen). All samples were acquired on a Gallios flow cytometer (Beckman Coulter). Data were analyzed using the software Kaluza Flow Cytometry Analysis v1.1 (Beckman Coulter).

The statistical tests used are described in the figure legends along with descriptions of measures of central tendency and dispersion as well as correction for multiple testing. The α value was set to 0.05 throughout the study. The software used was Prism 5 (GraphPad Software).

The collection and use of human samples were approved by the Regional Committee of Ethics (Regionala Etikprövningsnämnden, Stockholm and Uppsala) and were performed in accordance with the declaration of Helsinki. Written informed consent was obtained from each human subject that contributed samples for this study. All procedures performed on animals in this study were approved by the local ethics committee (Stockholms norra djurförsöksetiska nämnd).

IL22RA2 is the nearest gene, and thus candidate gene, for the MS risk locus on chromosome 6 identified by an observed case-control segregation of allele frequencies of SNP rs17066096 (3–5) (Supplemental Fig. 1). We hypothesized that rs17066096 genotype influences MS risk by altering transcription of IL22RA2 and set out to investigate this in human peripheral blood cells. No eQTLs for this variant are indicated in the Genotype-Tissue Expression portal database (gtexportal.org). Although transcription of IL22 and IL22RA1 was detected in human peripheral blood cells, IL22RA2 transcription was not (Fig. 1A). However, in the Genotype-Tissue Expression data, occasional whole blood samples have measurable quantities of IL22RA2, although median expression is 0 (gtexportal.org). To study the link between rs17066096 genotype and expression levels of IL22RA2, we therefore first had to set up a suitable expression system in vitro. This has the added advantage of enabling us to skew the cells with different stimuli, knowing that the effect of an SNP on transcription may appear only under certain conditions (20). IL22RA2 is known to be secreted by APCs in tissues (18, 21). Therefore, peripheral blood monocytes were isolated from healthy individuals and differentiated into APCs with similarities to macrophages and DCs. Culturing the monocytes for 6 d with IL-4 or the combination of GM-CSF and IL-4 induced prominent expression of IL22RA2 (Fig. 1B). Further, considerable amplification of IL22RA2 expression was achieved when the retinoic acid receptor agonist AM580 was included from the start of the differentiation protocol, with a maximum effect in combination with GM-CSF and IL-4. Similar results have been reported by others (21). The effect of AM580 on IL22RA2 expression in GM-CSF– and IL-4–differentiated monocytes was dose dependent with a maximum around 10 nM (Supplemental Fig. 2A). It has been shown previously that stimulation in vitro with LPS decreases expression of IL22RA2 in monocyte-derived DCs (16). We sought to confirm this and contrast it with the effects of signaling via the other TLR. Again, using monocytes differentiated with GM-CSF and IL-4 as baseline, dramatic reduction of IL22RA2 expression was achieved by LPS when added during the final 24 h of differentiation (Fig. 1C), even in the context of AM580-differentiated cells (Fig. 1D). Varying degrees of reduction were also seen by the other tested TLR ligands (Fig. 1C). In summary, IL22RA2 mRNA is not detected in circulating cells but can be highly induced in vitro by GM-CSF, IL-4, and AM580, resulting in a phenotype that has similarities with tissue APCs. Addition of established APC activation stimuli, TLR ligation, results in reduced IL22RA2 expression.

Next, we set out to investigate the impact of MS-associated SNP rs17066096 on IL22RA2 transcription. To avoid confounding factors such as severity of disease and inflammation, disease stage and type, as well as the wide range of different MS treatments, we chose to proceed with samples from healthy individuals. We collected blood samples from healthy donors, all preselected based on the rs17066096 genotype. Preselection was necessitated because the population frequency of the risk GG genotype is only ∼5%. Individuals homozygous for each allele were included. Following mononuclear cell extraction, the monocytes were isolated and then frozen in medium containing DMSO for later use. When the collection phase was completed, cells from all individuals were thawed and differentiated in parallel. The protocol used for these experiments involved differentiation of monocytes in the presence of GM-CSF and IL-4 for 6 d to induce a phenotype similar to DCs (Fig. 1E). The purpose of using so-called monocyte-derived DCs in this study is not to draw conclusions about the behavior of bona fide DCs in vivo but rather to investigate the potential of this genetic variant to influence gene expression in the monocyte lineage after exposure to plausible physiological stimuli. IL22RA2 expression in relation to rs17066096 genotype was assessed at the end of the experiment. In addition to baseline expression in the monocyte-derived DCs, we also included three more conditions to study the effect of rs17066096 genotype on IL22RA2 expression in DCs differentiated in the presence of AM580 and those that also had been matured with LPS stimulation. IL22RA2 expression was significantly higher in cells from the individuals carrying the MS-associated genotype rs17066096^GG when cultured with 1 nM of AM580 without or with 24 h of LPS stimulation (Fig. 1F). This is consistent with the MS risk SNP rs17066096 tagging an eQTL that under specific circumstances affects expression of nearby gene IL22RA2.

Although IL22RA2 transcription is not detectable in circulating cells, we detected the protein product in both plasma and CSF by ELISA. In vitro, we only saw a significant effect of rs17066096 genotype on IL22RA2 expression when the cells had been conditioned with retinoic acid receptor agonist AM580. Accordingly, no significant associations to rs17066096 genotype were observed in relation to levels of IL-22BP (Fig. 1G). Next, we sought to link IL-22BP levels in plasma and CSF with measures of MS severity, so we used the Expanded Disability Status Scale and lesion number on magnetic resonance imaging. In our database, lesion number is categorized on a 5-level scale (0, 1–2, 3–5, 6–8, and 9 or more lesions), with the highest level being more common than the other levels combined. For this reason, a cut-off was made at nine lesions to define low and high lesion load. Patients with high lesion load had significantly higher levels of IL-22BP in the CSF (Fig. 1H). No significant correlation between IL-22BP in CSF and age or Expanded Disability Status Scale was observed (Supplemental Fig. 1B), suggesting that IL-22BP covaries with inflammation rather than neurodegeneration.

To generate hypotheses regarding the role of the IL-22 system in autoimmune neuroinflammation, we studied how the expression of its components varied over the course of MS model EAE. The EAE model in C57BL/6 mice is widely used to study pathophysiological mechanisms relevant for MS. However, MOG-induced EAE in DA rats has some distinct advantages over the standard mouse model. Most notably, it requires a milder adjuvant for disease induction, presents with a relapsing remitting disease course similar to MS, and the larger animal size makes it easier to sample CSF for analysis (22, 23). Moreover, we have previously demonstrated a similar association of the homologous Il22ra2 gene with risk and severity of rat EAE (17). Samples were collected at fixed intervals before onset and at clinically predefined stages after onset of disease. In these experiments, onset of EAE was relatively late, after day 15. Tissue samples from the inguinal lymph nodes, which drain the site of MOG immunization, were collected from DA rats as well as PVG rats that are resistant under the same induction protocol. Expression of the Il22 gene increased moderately in both strains but more so in the DA strain, with a maximum around day 7, which corresponds to the peak of immune cell activation (Fig. 2A). The expression pattern of the Il22ra2 gene was essentially inverted compared with that of Il22. Low levels of Il22ra1 were detected, which increased slightly over time. Next, draining lymph nodes were collected from DA rats that were untreated (day 0) or 7 d after immunization, followed by enzymatic digestion and isolation of basic cell types. The Il22 gene mRNA was detected in all cell types (Fig. 2B). Interestingly, cell samples that had been depleted of most T cells, B cells, CD11b/c⁺ cells, and OX-62⁺ DCs (called marker negative in this article) had the highest expression of Il22. However, the marker-negative cells are relatively few, and their contribution to the total amount of IL-22 is therefore likely low. The increased Il22 expression seen in Fig. 2A is likely to primarily reflect the recruitment of IL-22–producing cells rather than increased production from resident cells. Consistent with the human data in Fig. 1, we observed the highest expression of the Il22ra2 transcription in APCs and a reduction upon activation. Interestingly, the IL-22R subunit Il22ra1, which is often reported as absent on immune cells, became detectable in one T cell–enriched sample 7 d after immunization.

The expression and role of the IL-22 system components in the CNS is even less well defined. In samples collected in a similar manner, transcripts and protein levels of IL-22 were transiently increased in the spinal cord and CSF during the onset of disease (Fig. 2C). We observed a slight increase in Il22ra2 transcription in the spinal cord during all stages of disease compared with naive rats, which was positively correlated with a trend toward increased CSF IL-22BP levels at the peak of disease that decreased as motor function was regained. IL-22R is not possible to measure by ELISA, but transcription of the receptor subunit Il22ra1 was detected in the spinal cord with a trend toward increased expression throughout the disease course. To further characterize the cellular sources of Il22, Il22ra2, and Il22ra1 in the CNS, their expression was determined in mouse brain samples. The choice of species was motivated by better availability of cell-specific markers for sorting in the mouse. Samples from naive brains and brains taken 29 d after EAE induction were dissociated and enriched for astrocytes, oligodendroglia, microglia, or neurons. Interestingly, Il22ra2 was detected in CD11b⁺ cells enriched for microglia (Supplemental Fig. 3A). Moreover, Il22ra1 mRNA was detected in O4⁺ oligodendroglia and, to a lesser extent, in astrocytes. IL-22 has a well-documented tissue-protective role in several organs outside the CNS (24). A report has also shown that human primary astrocytes exhibit less stress-induced apoptosis if IL-22 is added in the medium (25). In light of this and our previous data, we hypothesized that IL-22 may have a proliferative or protective effect on these glial cells. To test this, rat oligodendrocyte progenitor cells were differentiated in vitro with the addition of recombinant rat IL-22, but no effect on differentiation or cell numbers was observed (Supplemental Fig. 3C). In the same assay, astrocyte density could be assessed, but no effect was seen by IL-22 stimulation. We also treated mouse astrocytes with a panel of stressors with or without addition of IL-22 followed by Annexin V staining, but no effect on apoptosis was observed (Supplemental Fig. 3D). In summary, the expression patterns of IL-22 and IL-22BP are dynamically regulated over the course of EAE in rats, both in the lymph nodes where the immune response is initiated as well as in the CNS. IL-22R can be detected in both compartments but, contrary to Perriard et al (25), we could not demonstrate a trophic effect of IL-22 on astrocytes or on oligodendrocytes.

The human data suggest that genetically determined changes in IL22RA2 levels influence susceptibility and possibly also severity of MS (Fig. 1). We therefore hypothesized that moderate changes in Il22ra2 expression will influence susceptibility to or severity of rat EAE by acting in the draining lymph node or in the CNS. For this purpose, we developed an inducible Il22ra2 knockdown rat strain in which administration of doxycycline in the drinking water results in a dose-dependent decrease in Il22ra2 expression, mediated by short hairpin RNA (Fig. 3A, 3B, Supplemental Fig. 4A, 4B). With this approach, the rats have normal Il22ra2 expression until the start of doxycycline administration, thus avoiding potential compensatory changes during development. Starting doxycycline treatment at 9 d after EAE induction, which knocks down Il22ra2 expression after the priming events have taken place, was partially protective in this model but required 500 mg/l of doxycycline (Fig. 3C). However, knocking down Il22ra2 expression before immunization had a dose-dependent inhibitory effect on EAE, reducing severity with 4 and 20 mg/l of doxycycline and making the transgenic rats completely resistant to disease with 100 mg/l (Fig. 3D), a dose that had no effect when administration started 9 d after immunization (data not shown). These findings further support a role of IL-22BP during priming in the periphery rather than during the effector phase or the response in the CNS.

The data so far indicate potential roles for IL-22BP in both the periphery and in the target organ. However, expression of IL-22BP is much higher in secondary lymphoid tissue (Supplemental Fig. 4B), and the effect of knocking down gene expression before immune cell activation has taken place is dramatic (Fig. 3D). Moreover, no effect of IL-22 on cultured glial cells could be demonstrated (Supplemental Fig. 3C, 3D). Keeping in mind that an effect of IL-22 in CNS during EAE cannot be formally excluded, we now focus on the periphery in which the IL-22 system clearly is involved. Hypothesizing an effect of IL-22BP on the initiation of the immune response in EAE, we focused on T cells in the draining lymph nodes 7 d after immunization. This is further justified by our observations of IL22RA1 expression in human T cells ex vivo (Fig. 1A) and upregulation of Il22ra1 expression in draining lymph nodes 7 d after MOG immunization in rats (Fig. 2A, 2B). In experiments performed as in Fig. 3D, with 100 mg/l of doxycycline, the transgenic rats have significantly smaller lymph nodes 7 d after immunization (Supplemental Fig. 4C, 4D), consistent with a facilitating role of IL-22BP on immune cell priming/expansion. When assessing T cell polarization in draining lymph nodes 7 d after MOG immunization, we observed a significant decrease in Ag-specific IFN-γ⁺ T cells (Fig. 4A). This suggests that IL-22 acts to dampen the immune response, with a prominent effect on Th1 polarization. We set out to investigate this in human T cells and performed polyclonal stimulation of human peripheral blood CD4 or CD8 T cells with Abs for CD3 and CD28. We observed upregulation of IL22 transcription in CD4 T cells after 24 h of stimulation (Fig. 4B). IL22RA1 was detected ex vivo in both subsets, but no significant change was induced after 24 h of polyclonal stimulation (Fig. 4C). To test if additional IL-22 stimulation has an effect on IFN-γ production, we activated CD4 or CD8 T cells for 3 d and stimulated them simultaneously with recombinant human IL-22. A statistically significant difference was seen with higher IFN-γ production in CD4 T cells that had not been activated in the presence of IL-22 (Fig. 4D). A similar trend was observed in cultures of CD8 T cells.

Our data suggest that carriers of the risk genotype rs17066096^G have higher expression of IL-22 antagonist IL-22BP. Furthermore, in this study, we show rat and human data implying that IL-22 has the potential of suppressing the Th1 signature cytokine IFN-γ. Consequently, we hypothesized that carriers of rs17066096^G would, via decreased availability of IL-22, exhibit a higher capacity for IFN-γ production compared with noncarriers. To investigate this, we used samples available to us from cultures of PBMCs from 84 patients with MS or other neurologic diseases with known rs17066096 genotype. These cells had been stimulated for 3.5 h in medium only or with the addition of LPS, thus approximating physiological activation of memory T cells via innate immune cells. We determined the levels of IFNG gene expression in these samples and observed significantly higher levels in carriers of the risk genotype (Fig. 4E). This is consistent with the role of IFN-γ as an activator of innate immune cells, the effector arm of the Th1 immune response, and also a documented pathogenic role of IFN-γ in MS (26).

Next, we asked if the disease-driving effect of IL-22BP on EAE that we observed in this study in rats, and have previously reported in mice (18), is mediated by IL-22. To investigate this, we performed EAE experiments with mice in which either Il22 or Il22ra2 was deleted or in which both genes were deleted. We observed less-severe disease in Il22ra2^−/− mice compared with wild-type mice (Fig. 5A). Interestingly, heterozygous deletion of Il22ra2 was sufficient to achieve the full protective effect. However, the protective effect of Il22ra2 deletion was completely lost when the Il22 gene was also deleted (Fig. 5B). The absence of the Il22 gene alone had no effect on EAE score, which has been reported previously by others (27). There was no difference in day of disease onset when comparing the four strains. Infiltrating immune cells were isolated from whole spinal cord of perfused mice on day 29 of EAE and characterized by flow cytometry. Il22ra2^−/− mice had fewer infiltrating T cells compared with wild-type mice as well as the double knockout mice (Fig. 5C). Interestingly, Il22^−/− mice had more infiltration of neutrophils and a trend toward more T cells and inflammatory monocytes, compared with wild-type mice.

We were first, to our knowledge, to link IL-22BP to neuroinflammation in a study of rat and human natural genetic variation in this locus, demonstrating effects on EAE severity as well as MS susceptibility and severity (17). The protective genetic variant in the rat strain used in that study (17) conferred lower expression of IL-22BP, making it the candidate gene for the effects seen in EAE. We then went on to pinpoint the effect on EAE to the actual gene using IL-22BP–deleted mice (18), which have less severe disease. We have since then shown, in three large-scale studies, that the G allele of SNP rs17066096, located 14 kb downstream of the gene for IL-22BP, is associated with higher MS susceptibility (3–5). The combined p value of 9.26 × 10⁻³⁶ makes this one of the strongest MS-associated non-HLA variants (5). It was not known how this increased risk is mediated, but based on the data that we have published on mice and rats, we hypothesized that the MS-associated G allele would be associated with higher expression of IL-22BP, thereby mediating the pathogenic effect.

Overall, there are very few examples of uncovered mechanisms behind genetic associations to complex diseases in the literature, especially when considering noncoding variants. However, several successful studies have, like we have done in this study, used samples from healthy individuals (28–34). In the current study, we first defined an in vitro system in which IL-22BP gene expression can be studied, necessitated by the fact that mRNA levels are not readily detected in circulating cells. Just like rodent tissue DCs (18), these human in vitro differentiated monocytes produced high levels of IL-22BP, providing a system to study the role of genetic variation. This approach has also been used successfully by others (30). In fact, emerging data suggest that noncoding variants often require specific context to exert their effect, and this has recently also been suggested for autoimmune disease risk genes, including MS (20, 35). When studying IL-22BP expression using this approach, we observed context-dependent differences relative to rs17066096 genotype, consistent with an eQTL. Furthermore, CSF IL-22BP levels were higher in MS patients with more severe neuroinflammation. The pathogenic effect of IL-22BP in neuroinflammation was demonstrated in mouse and rat EAE, which correlated with a reduction in IFN-γ production from draining lymph node encephalitogenic T cells. Moreover, rat lymph node cells had upregulated the membrane-bound receptor 7 d after EAE induction. Translating the findings back to human cells, we demonstrate the ability of IL-22 to limit T cell IFN-γ production and provide further evidence for this effect as well as the eQTL in an experiment using cells from a large number of patients with MS and other neurologic diseases.

In an early report, IL-22–deleted mice were shown to have an unaltered clinical and histopathological presentation of EAE compared with wild-type mice (27). In contrast, we show that IL-22BP–deleted mice have less severe EAE compared with wild-type mice. One can speculate that the presence of IL-22BP, and its efficient IL-22 antagonism, masks the potential of IL-22 to significantly influence EAE disease course when comparing with IL-22–deleted mice. Another possible explanation is that the effect of IL-22BP is mediated independently of IL-22. In this study, we addressed this by conducting a series of EAE experiments with wild-type, IL-22–deleted, and IL-22BP–deleted mice as well as mice in which both genes are deleted. We could conclude that the reduced severity of disease in IL-22BP–deleted mice is an effect of disinhibited IL-22 signaling. Moreover, like the previous study of EAE in IL-22–deleted mice (27), we saw no effect on EAE score, but when analyzing the infiltrating cells, we detected more neutrophils and a trend toward more T cells. This further supports a protective role for IL-22 in neuroinflammation, but it also shows that this effect is to a large degree restrained in wild-type mice because of the presence of IL-22BP. Another interesting aspect of the mouse EAE experiments is that heterozygous deletion of the IL-22BP gene is sufficient to achieve the full protective effect, analogous with the subtle knockdown of IL-22BP in the rat EAE experiments, which serves as indirect evidence that the moderate-sized modulation of human IL-22BP expression by the risk variant is clinically relevant.

The inverse patterns of IL-22 and IL-22BP expression in the lymph node that we observed during the first days after EAE induction are consistent with the existence of common upstream signaling events regulating transcription in a manner that facilitates IL-22 signaling, that is, as IL-22 increases, the homeostatic expression of antagonist molecule IL-22BP is dampened. These events could be interpreted as a physiological upregulation of IL-22 signaling as a consequence of tissue damage from the s.c. injection of Ag emulsified with adjuvant, with a concomitant decrease in IL-22BP to facilitate the tissue-regenerative actions of the cytokine. Alternatively, IL-22 is induced as a physiological feedback inhibition of the events during immune cell priming to limit excessive activation. Our data suggest that the latter is at least plausible, considering the anti-inflammatory effect on rat and human T cells. Further support for this interpretation can be found in a study of experimental autoimmune uveitis, in which administration of IL-22 on day 4 and 8 after immunization results in milder disease (36). They report that in vitro stimulation of T cells from IL-22–treated mice results in decreased IL-17 and IFN-γ secretion as well as an increased proportion of Foxp3⁺ T cells. Results that are essentially similar to what we show in this study but using a different experimental intervention, disease model, and species. Moreover, they also detect upregulation of IL-22R in lymphoid organs after immunization but localize it to APCs rather than T cells. However, the method of isolation in that report was adherence to nylon wool columns, which is known to alter the activation state and marker expression of T cells (37).

In this study, we show IL-22BP expression in the CNS. The major CNS cell types were sorted from untreated neonatal wild-type mice, and IL-22BP expression was present and restricted to CD11b⁺ cells. Moreover, IL-22R was detected in mouse brain samples enriched for O4⁺ oligodendroglia and to a lesser extent in astrocytes. Two previous studies have addressed IL-22 signaling on glial cells, arriving at somewhat contradictory conclusions. Human astrocytes cultured under stress are protected by IL-22, whereas mouse oligodendrocytes are more likely to go into apoptosis under similar conditions (25, 38). In our assays, IL-22 stimulation of mouse and rat glial cells had no effect, and knocking down IL-22BP after immune cell activation had no strong effect on disease course of EAE in rats. IL-22BP expressed in the periphery is thus likely to have a more decisive role in the context of EAE but not necessarily in MS.

In summary, although the knowledge of the genetic architecture of MS susceptibility has expanded immensely in the last few years, real clinical benefit from these efforts will require functional annotations of the associated sequence variants. In this study, we have focused on one of the strongest MS-associated non-HLA variants as well as the nearby localized gene for IL-22BP. We have presented evidence for a protective role of IL-22 in the context of MS, thus advancing our understanding of this enigmatic disease and suggesting IL-22BP as a novel MS drug target.

The authors have no financial conflicts of interest.