|Home | About | Journals | Submit | Contact Us | Français|
Interferon alpha (IFNα) is known to play a key role in autoimmunity, but the mechanisms are uncertain. Although the induction of autoimmunity by IFNα is consistent with primarily immuno-modulatory effects, the high frequency of non-autoimmune inflammation suggests other mechanisms. We used thyroiditis as a model to dissect these possibilities. IFNα treatment of cultured thyrocytes increased expression of thyroid differentiation markers, thyroglobulin (Tg), thyroid stimulating hormone receptor (TSHR), thyroid peroxidase (TPO), and sodium iodide transporter (NIS). RNAseq analysis demonstrated that pathways of antigen presentation, pattern recognition receptors, and cytokines/chemokines were also stimulated. These changes were associated with markedly increased non-apoptotic thyroid cell death suggesting direct toxicity. To corroborate these in vitro findings we created transgenic mice with thyroid specific overexpression of IFNα under control of the Tg promoter. Transgenic mice developed marked inflammatory thyroid destruction associated with immune cell infiltration of thyroid and surrounding tissues leading to profound hypothyroidism, findings consistent with our in vitro results. In addition, transgenic mice thyroids showed upregulation of pathways similar to those observed in cultured thryocytes. In particular, expression of granzyme B, CXCL10, a subset of the TRIM (tripartite motif containing) family and other genes involved in recruitment of bystander cytotoxic immune responses were increased. Pathways associated with apoptosis and autophagy were not induced. Taken together, our data demonstrate that the induction of tissue inflammation and autoimmunity by IFNα involves direct tissue toxic effects as well as provocation of destructive bystander immune responses.
Interferon alpha (IFNα), a critical cytokine in the host immune response to viral infections and tumors, has been extensively used as a therapeutic agent for a growing array of diseases, ranging from malignancies, such as hairy cell leukemia (1), to viral infections, notably chronic Hepatitis C virus (HVC) infection (2). IFNα binds to the type I interferon receptor, a transmembrane glycoprotein dimer with cytoplasmic domains. Binding induces a signaling cascade mediated primarily through the JAK-STAT, Crk, IRS, and MAP kinase pathways (3). Yet, the full scope of the actions of INFα, either produced naturally or by therapeutic delivery, remains poorly understood. Emblematic of this understanding gap is the disparate array of notable side effects associated with IFNα therapy. These include influenza-like symptoms, hematological perturbations, neuropsychiatric signs, and thyroid disease. Individually and/or cumulatively these side effects can lead to dose reductions in up to 40% of patients and drug discontinuation in up to 14% of patients (4–7).
IFNα also plays a major role in the development of autoimmunity. Activation of IFNα pathways has been implicated in a number of autoimmune diseases, most notably systemic lupus erythematosus (SLE) (8). The association IFNα therapy with the development autoimmune thyroiditis has, in particular, supported a role for IFNα in provoking autoimmune responses (9,10). However, the mechanisms by which IFNα induces autoimmunity are still unknown. Conceptually, the induction of autoimmunity is consistent with various immuno-modulatory effects of IFNα, such as activation of cytokines and adhesion molecules, and, of cardinal importance, stimulation of MHC class I antigen expression (11,12). Moreover, IFNα shifts the immune response to a Th1 pattern (13–15), resulting in the production of IFN-γ and IL-12, two potent proinflammatory cytokines (15). IFNα can induce the release of other cytokines, such as IL-6, a cytokine that has been associated with autoimmune thyroiditis (16).
Although the immune effects of IFNα provide an attractive basis for the induction of autoimmunity by IFNα, the high frequency of non-autoimmune tissue inflammatory conditions, most notably destructive thyroiditis in many cases of thyroiditis associated with IFNα therapy (10), suggests that IFNα can induce pathologic processes also by atypical immune and/or non-immune mechanisms. We, therefore, hypothesized that IFNα triggers autoimmunity in genetically susceptible individuals by a combination of direct tissue toxicity and immune modulation. To test this hypothesis we used interferon induced thyroiditis (IIT) as a model.
IIT is a well recognized syndrome associated with IFNα therapy that was first described in 1983 in patients treated with IFNα for carcinoid tumors (17,18) and breast cancer (19). Subsequently, numerous studies reported a high incidence of IIT in patients treated with IFNα, mostly for hepatitis C infection (10,20). Intriguingly, IIT can manifest either as autoimmune or non-autoimmune thyroiditis, an observation suggesting that IFNα induces tissue inflammation by immune as well as non-immune mechanisms (9–11). Our data, combining in vitro and in vivo studies, support the hypothesis that IFNα has direct tissue toxic effects, most notably the induction of thyroid cell necrosis. Moreover, we show that IFNα provokes a clear stimulation of an immune-regulated and destructive inflammatory bystander response which likely triggers tissue-specific autoimmunity in a genetically susceptible host.
Dulbecco’s Modified Eagle’s Medium (DMEM) and penicillin-streptomycin were purchased from Fisher Scientific (Pittsburgh, PA). 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, a tetrazole (MTT), Coon’s modification of Ham’s F12 media, thyroid-stimulating hormone, insulin, apotransferrin, and hydrocortisone were purchased from Sigma (St. Louis, MO). TRIzol solution and fetal bovine serum were purchased from Invitrogen (Carlsbad, CA). StrataScript QPCR cDNA Synthesis Kit and Brilliant SYBR Green QPCR Reagents were purchased from Stratagene (La Jolla, CA). Mouse anti-human TSH Receptor antibody and Fluorescein isothiocyanate (FITC)-conjugated rabbit anti-mouse IgG were purchased from Serotec (Raleigh, NC). FITC-conjugated mouse anti-rat MHC class I monoclonal antibody and mouse anti-human beta actin monoclonal antibody were purchased from Abcam (Cambridge, MA). FITC-conjugated nonspecific mouse immunoglobulin G1 (IgG1) control was purchased from BD Biosciences Pharmingen (San Jose, CA). Phycoerythrin (PE)-conjugated goat anti-mouse IgG antibody, normal mouse IgG1 and FCM wash buffer were purchased from Santa Cruz (Santa Cruz, CA). Purified hamster anti-mouse CD3e, biotin rat anti-mouse CD45R/B220 and purified rat anti-mouse F4/80-like receptor monoclonal antibodies were purchased from BD Biosciences Pharmingen (San Jose, California).
Creation of the TG mice and all mouse studies were reviewed and approved by the University of Cincinnati and Mount Sinai Institutional Animal Care and Use Committees. The mouse IFNα (mIFNα) cDNA, kindly provided by Dr. T. Michiels (University of Louvain, Brussels, Belgium), was digested by BamHI and XhoI into a 0.6 kb mIFNα fragment and ligated into a pSG5 vector containing the rabbit beta-globin second intron and SV40 poly A tail. The pSG5-mIFN construct was digested with StuI and SalI and ligated into a pBluescriptSK (+) vector containing the bovine Tg (bTg) promoter (pBSK-bTg) (kindly provided by Dr. J. Fagin MSKCC, NY) at the EcoRV and SalI sites. The final product (designated bTg-mIFNα, Figure 3, Panel A) was verified by direct sequencing.
The plasmid containing the bTg-mIFNa construct was digested with SacI/KpmI, generating a 3.4 Kb fragment (Figure 3, Panel A). Purified transgene DNA (3.4 Kb) was resuspended in 5mM Tris-HCl, pH 7.4/0.15mM EDTA and given to the UC Transgenic Core facility for microinjection. C57BL/6JxCBA/J F1 hybrid mice were superovulated and mated with fertile C57BL/6 males. Single cell embryos were harvested and pronuclei microinjected with 1–2 pl of DNA solution. Embryos surviving microinjection were reimplanted into pseudopregnant C57BL/6J females. Founders were identified initially by PCR screening of genomic DNA with TG-specific primers. Forward primer (5′-GTTTGGGGACCCTTGATTGTTCTT-3′) and reverse primer (5′-AGGGGCCCGGTTTGGACTC-3′). Reference control was TSH receptor gene; forward primer (5′-GTAACTCACTCATGCAAAGT-3′) and reverse primer (5′-TCCTCAAAGATGCTCATTAG-3′). In order to verify the integrity of the transgene Southern blotting of XbaI/SalI digested genomic DNA (size 3.4 Kb), from tail biopsies was performed using standard techniques (Figure 3, Panel C).
Human thyroid primary cells were prepared from fresh, non-cancerous thyroid tissue adjacent to thyroid tumors that were removed at surgery. Tissues were obtained from the University of Cincinnati (UC) tissue bank. The use of deidentified anonymous human thyroid was approved by the UC institutional review board. Tissue was minced and incubated in 200 U/ml of collagenase solution for 1 hour at 37°C. Cells were harvested and cultured in DMEM supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 μg/ml) (P0). Cells were passaged at 1:2 dilution and cultured until confluent (P1). P1 or P2 cells were used for experiments. Cells were confirmed to be thyroid cells by Western Blot analysis for thyroglobulin (data not shown).
The well-differentiated, nontransformed rat thyroid cell line, PCCL3, was kindly provided by Dr. James Fagin (Memorial Sloan-Kettering Cancer Center, NY) and propagated in H4 complete medium, which consisted of Coon’s modification of Ham’s F12 media containing 5% FBS, glutamine (286 μg/ml), apotransferrin (5 μg/ml), hydrocortisone (10 nmol/l), insulin (10 μg/ml), thyroid-stimulating hormone (10 mIU/ml), penicillin, and streptomycin. Cells were cultured at 37°C in 5% CO2; medium was replaced every 48 hours. Cells were counted with a Z1 Coulter Counter Cell and Particle Counter (Beckman Coulter, Inc. Fullerton, CA).
Total cellular RNA was extracted from rat tissues and cells using TRIzol reagent (21). For cell culture and RNAseq confirmation QPCR experiments, 5 μg of RNA was used to synthesize cDNA by StrataScript QPCR cDNA Synthesis kit. mRNA levels were measured by quantitative real-time PCR (Q-PCR) using a SmartCycler System (Cepheid, Sunnyvale, CA) using the primers shown in Supplemental Table 3 and SYBR green for detection of amplicon. For Q-PCR determination of mRNA expression levels of different genes cDNA was obtained from 1 ug of RNA using the Superscript III first strand cDNA synthesis kit (Invitrogen, Carlsbad, CA) and RT-PCR was performed using an ABI 7300 machine and Taqman probe system (Applied Biosystems [ABI], Foster City, CA). The IFNα gene does not contain introns (22) and, therefore, intron spanning primers could not be provided by ABI. Accordingly, after RNA extraction and before cDNA synthesis, the RNA preparation was treated with DNAse using the reagent kit, Turbo DNA-free (Ambion, Austin, TX) according instructions and then reverse transcribed. The primers used to measure mouse IFNα levels were obtained from the inventoried set available from ABI (mouse interferon α 2, Assay ID: Mm00833961_s1). The reference gene was mouse beta-actin (mouse ACTB, VIC-MGB, ID: 4362341E-0708008). Other RNAs assayed using inventoried ABI primer/probes mixes included the following: mouse OAS1a (ID: 00836412.m1); mouse GranzymeB (Mn00442874.m1); mouse TRIM21 (Mn00447364.m1); mouse CXCL10 (Mn9999072.m1); rat CXCL10 (Rn00594648.m1); rat granzyme B (Rn00821752.gi); human OAS1 (Hs00973635.A1); human TRIM21, (Hs001726.m1); and human granzyme B, (Hs00188051.m1). Reactions were performed in triplicate. For time and dose-course experiments normalized expression was determined by applying the Ct values for target and reference genes to the Q-Gene program (23) (Figures 8–9). For the Q-PCR experiments performed to confirm the RNAseq results (Supplemental Table 2) we calculated the −ΔΔCt as follows: Ct values of tested genes were first normalized to the Ct values of β-actin to obtain a ΔCt (i.e. Ct of tested gene-Ct of β-actin). The ΔCt value of the wild type (WT) mice was then subtracted from the ΔCt value of the transgenic (TG) mice for the same gene. The obtained value of ΔΔCt is equal to the negative of the logarithm base 2 of the ratio of mRNA levels in the TG mice to of mRNA levels in the WT mice, and therefore, the − ΔΔCt value is shown in Supplemental Table 2 for consistency with the RNAseq results.
Transient transfection of cell lines was performed using the Lipofectamin 2000 kit according to the manufacturer’s instructions.
PCCL3 rat thyroid cells were seeded into 6 cm dishes. Cells were harvested in phosphate-buffered saline (PBS), washed with PBS, and resuspended in FCM wash buffer supplemented with 0.02% sodium azide. For detection of MHC class I, cells (5 × 105) were incubated for 30 min at 4°C with 10 ul of FITC-conjugated mouse anti-rat MHC class I monoclonal antibody or FITC-conjugated nonspecific IgG1 control and washed twice. For TSHR, 5 × 105 cells were incubated with 1 ug of mouse anti-human TSH Receptor antibody (or control IgG1) for 30 min at 4°C, washed twice with FCM wash buffer, followed by incubation with PE-conjugated goat anti-mouse IgG (1:200). Antibody binding was quantified by flow cytometry (mean fluorescence intensity [MFI]), using a COULTER EPICS XL-MCL Flow Cytometer (Beckman Coulter, Inc. Fullerton, CA).
PCCL3 thyroid cells or human thyroid primary cells were plated into ninety six-well plates at 3 × 103 or 1 × 104 cells per well and treated with various concentrations of IFNa for 48 hours. Cell viability was measured by MTT assay as previously described (24).
PCCL3 rat thyroid cells or human thyroid primary cells were cultured in 6 cm dishes and treated with IFNa for 48 hours. Cells were harvested and washed in cold PBS. Necrotic and apoptosis cells were measured by a Vybrant apoptosis assay kit #3 (Invitrogen; Carlsbad, CA).
Human thyroid primary cells were grown as described above and exposed to IFNα 5000 u/ml for 12 and 24 hours. Total RNA was purified from cells using TRIzol, DNAse treated, and electrophoresed to confirm integrity. For the analysis of mouse thyroid tissues we used a cohort of transgenic animals, high expressing IFNα line 91 in an approximately 75% C57Bl6/25%CBA/J background, that were given T4 supplementation until 6 months of age, and were then taken off T4 replacement for approximately 2 months. The animals were sacrificed at 8 months of age. The atrophied but dissectible thyroids were removed and pooled for RNA extraction (total of 10 animals). Control thyroids from litter mate controls of normal size were also pooled for analysis (n= 8). Because of constraints related to the small size of the TG thyroids and the cost of the RNAseq, we performed RNAseq of pooled samples. Therefore, to increase the power of our analyses the cutoff for significance was adjusted to be robust and the threshold number of sequences required for significance was increased (see below). The RNA from the thyroids was extracted using the RNeasy min kit (Qiagen, Inc, Valencia, CA), DNAse treated (RNase-free DNase (Qiagen), analyzed for integrity by agarose gel electrophoresis.
RNAseq is a method of transcriptome analysis that consists of sequencing a cDNA library by high throughput next generation sequencing. The number of reads aligning to a specific gene sequence is proportional to the abundance of that gene in the sample from which the cDNA library was prepared. We used the mRNA-Seq Sample Preparation Kit from Illumina (San Diego, CA) according to the manufacturer’s recommendation. Briefly, mRNA was extracted from 2ug of total RNA using oligo-dT magnetic beads and fragmented at high temperature using divalent cations. A cDNA library compatible with the Illumina next generation sequencing technology was then prepared from the fragmented mRNA by reverse transcription, second strand synthesis, and ligation of specific adapters. The amount of dsDNA in each library was accurately quantified by spectrofluorometric analysis using the Qbit system from Invitrogen (Carlsbad, CA) and diluted to a 10nM concentration. Next generation sequencing was performed on an Illumina Genome Analyzer IIx according to the manufacturer’s recommendations using the Single-Read Cluster Generation Kit v2 and the SBS Sequencing Kit v3. Image analysis and base calling was conducted in real-time by the SDS 2.5/RTA1.5 software. The reads with good quality were aligned to reference sequence databases of human (ucsc hg18) or mouse (mm9) genome, RefSeq exons, splicing junctions and contamination databases including ribosome and mitochondria sequences using BWA22 and the alignment files in SAM format were generated. After filtering reads mapped to contamination databases, the reads that were uniquely aligned to each exon and splicing-junction sites were extracted and then counted. The read count for each RefSeq transcript was calculated by combining the counts for exons and splicing junctions of corresponding transcript and also normalized to relative abundance in Fragments Per Kilobase of exon model per Million (FPKM) in order to compare transcription level among samples.
To compare the expression levels of transcripts across samples, the read counts of transcripts in each sample were normalized by leveling the total read counts in each sample to the maximum of the read counts in all samples. The read count data were then formatted into microarray-like data that could be analyzed using a variety of microarray statistical analysis tools. The differentially expressed transcripts were identified using M-A based random sampling method implemented in DEGseq package in BioConductor (http://bioconductor.org/packages/2.5/bioc/html/DEGseq.html). The transcripts were further filtered at > 2-fold change and a minimum read count of 50 in either condition.
The differentially expressed transcripts were subjected to pathway analysis by Ingenuity Pathway Analysis (IPA) system, version 8.6 (http://www.ingenuity.com/). The IPA program identifies biological networks and/or pathways representing interactions between the differentially expressed genes in the tested samples and/or with other genes in the database. The fold changes of these genes were converted to log2Ratio and then imported into the IPA tool along with gene symbols. Fisher’s exact test was used to calculate a p-value for the probability that a pathway was significantly enriched in input genes compared to the genome, and the pathways/networks were ranked by the p-values. The significance level and the percentages of up- or down-regulated genes of each pathway were summarized in bar-charts (Figure 6). The graphical presentation of gene-gene interactions and de-regulated genes for enriched pathways were visualized in Canonical Pathway Explorer (Figure 7).
TSH assay was performed by Drs. Samuel Refetoff and Roy Weiss (Division of Endocrinology, University of Chicago) as previously described (25). While normal range for adult mice is strain and gender specific for most strains TSH < 10 uU/ml is normal. The intra-assay and inter-assay coefficients of variations at the higher range of the assay are 10 and 24%, respectively. Total T4 levels were measured from blood spotted on filter paper using the neonatal T4 kit (Diagnostic Products Corporation, Los Angeles, CA) according to the manufacturer’s instructions. Normal range for total T4 was 2–8 ug/dl.
Following documentation of profound hypothyroidism in the high expressing IFNα transgenic line, the animals’ chow was changed to Modified PicoLab Rodent 20/250 ppm thyroid powder (TestDiet, Greenfield, IN).
The serum levels of mouse IFNα were determined by using a murine IFNα ELISA kit from PBL Biomedical Laboratories (Piscataway, NJ) according to the manufacturer’s instructions.
After sacrifice the thyroids were dissected in 2 ways: (1) Most thyroid specimens were dissected with the overlying strap muscles removed; (2) Nonthyroid hormone treated thyroid specimens from the two high-expressing IFNα lines were dissected with inner layer of overlying muscles unperturbed due to inflammatory changes in the surrounding tissue. Gross thyroid images were using a Leica E24 Stereoscopic microscope at variable magnification; digital images were captured using a Nikon Coolpix 950 camera. For routine hemotoxlin and eosin (H&E) staining, the trachea were cut above and below the area of the thyroid and the tissue was placed as a block into 10% formalin for at least 48 hours for tissue fixation followed by storage in 70% ethanol. After embedding in paraffin, standard 5 micron sections were obtained for staining. For immunocytochemistry, dissected tissues were embedded in OCT, sectioned, and placed onto slides. Prior to staining, slides were refixed in cold PBS containing 0.2% glutaraldehyde for 10 min. After staining, slides were washed in PBS, dehydrated through an ethanol series, and then coverslipped. Immunochemistry was performed by the Comparative Pathology Core at UC College of Medicine, Department of Pathology. Positive control for each of the three primary immune cell classes, CD3e (T cells), CD45R/B220 (B cells) and F4/80-like receptor (macrophages), were spleen sections processed identically and from the same group of animals used for immunocytochemistry. Serial dilutions of the primary antibody were performed with the spleen sections to optimize specific signal vs. background followed by simultaneous of spleen and thyroid tissue section on the same slides under identical conditions. Final concentrations of primary antibody were as follows: CD3e, 1:200; CD45R/B220, 1:5000; F4/80-like receptor, 1:200. Sections were visualized using an Olympus BX51 microscope, digital image captured using a Diagnostic Instruments digital camera, Model 74 Slider and image saved digitally using Spot Advanced Diagnostic Instruments software Windows version 4.6.
To ascertain whether IFNα can exert biological effects on thyroid cells, we first confirmed that the IFNα receptor (IFNαR) was expressed in the thyroid. 11 human tissues were analyzed by QPCR for IFNαR gene expression. Liver tissues showed the highest expression levels of IFNαR (26). Interestingly, thyroid tissues showed the second highest levels of expression of IFNαR among the tissues examined (Figure 1A). We next examined the effects of IFNα on the expression of the following thyroid specific genes: TSH receptor (TSHR), thyroglobulin (Tg), thyroid peroxidase (TPO), and sodium iodide symporter (NIS). We used the MHC Class I gene, known to be upregulated by IFNα, as positive control. PCCL3 cells (a differentiated rat thyroid cell line) were incubated with rat IFNα. Both dose-dependent responses and time-dependent responses were assessed. Time-course experiments performed with a constant IFNα concentration of 5000 U/ml showed a peak expression of TSHR mRNA at 24 hours that persisted through 48 hours (Figure 1B). The increase in TSHR was also apparent at the protein level as determined by FACS analysis. (Figure 1C). Similarly, the expression levels of Tg, TPO, and NIS mRNA were significantly increased at 24 hours by IFNα; Tg expression persisted even at the 48 hour time point, but the expression levels of TPO and NIS decreased to baseline by 48 hrs. (Figure 1D, E, F). To analyze the mechanism of increased expression of thyroid specific genes, promoter activities were determined by a luciferase assay for Tg, TPO, and NIS. PCCL3 cells were transfected with luciferase reporter plasmids, pTg-Luc, pTPO-Luc, pNIS-Luc to determine promoter activities of Tg, TPO and NIS, respectively. pGL4.10 was used as a negative control. Only the Tg promoter showed a significant increase in activity induced by IFNα (p<0.05) suggesting that TPO and NIS are upregulated by IFNα through other mechanisms. Finally, as expected, there was a dose-dependent and time-dependent increase in the expression of MHC class I mRNA when PCCL3 cells were incubated with rat IFNα (Figure 1G and H), as well as at the protein level (Figure 1I).
To test our hypothesis that IFNα is directly toxic to thyroid cells, we investigated the effect of IFNα on thyrocyte viability in cultured thyroid cells. A significant IFNα dose-dependent decrease in cell viability was seen in both PCCL3 rat and human thyroid cells in primary cultures with maximum cell death occurring at concentrations of 5000 U/ml of IFNα (Figure 2A and B). To determine whether IFNα-induced thyroid cell death was caused by apoptosis or necrosis, thyrocytes were treated with 5000 U/ml of IFNα for up to 48 hours and analyzed by the Vybrant apoptosis assay kit. Both PCCL3 and human thyroid cells (in primary culture) showed significant increase in necrotic cell numbers at 24 hours and 48 hours of IFNα treatment, but apoptotic cell numbers were unchanged (Figure 2C and D). mRNA levels of Caspase-3, (the major mediator of apoptosis) were also measured by QPCR. While a trend towards increased Caspase-3 levels was seen during IFNα treatment of PCCL 3 cells, this change was not statistically significant (Figure 2E). A third mechanism of cell death, autophagy (27,28) involving inhibition of the mTOR signaling pathway, (29) was tested by measuring S6K phosphorylation. There was no change of phosph-S6K levels on Western blot analysis after treatment with IFNα (Figure 2F), suggesting no role for autophagy in the observed cell death.
In order to examine the direct and indirect effects of IFNα on thyrocytes in vivo in a more direct physiologic manner we generated transgenic (TG) mice over-expressing IFNα in the thyroid (TG is our abbreviation for “transgenic” and Tg is the abbreviation for “thyroglobulin”). To generate the TG mice we used a construct in which the IFNα gene was placed under the control of the bovine Tg (bTg) promoter (Figure 3A). To confirm that immunoreactive IFNα could be produced by the construct, PCCL3 thyroid cells were transfected with the bTg-mIFNα construct and mIFNα protein levels were measured by ELISA. Significant amounts of mIFNα protein were secreted from PCCL3 cells after transfection with the bTg-mIFNα construct (Figure 3B). Transgenic mice generated as described in Methods, were screened by PCR (Figure 3C). PCR screening of the offspring identified 7 founders. 6 of the 7 founders (Animal Lines 91, 99, 22, 100, 30 and 2) appeared to have intact insertion of the bTg-mIFN-α transgene, as determined by Southern blot of F1 offspring, and were investigated further. Figure 3D shows a representative Southern analysis of two lines, 91 and 100. In order to confirm the tissue-specific and relative line expression of IFNα, we performed Q-PCR analysis of IFNα mRNA extracted from representative tissues. Figure 3E shows high levels of thyroidal expression of IFNα in an intermediate copy number line 100. Note the very low levels of expression in the other tissues tested.
Lines 91, 99 and 22, which demonstrated greater copy number by Southern analysis, exhibited reduced fertility and increased morbidity and mortality within the litters. Litter size ranged from approximately 3–8 in numbers with variable runting in the TG pups. Pups that survived to weaning but were runted initially proved to be invariably TG positive yet displayed catch-up growth on T4 supplemented chow. Figure 4A shows the pronounced runting with reduced body hair of the TG positive animals in a representative litter from Line 91 at day 14 of life. To assess for systemic toxicity of IFNα, we assayed serum from TG mice for IFNα by ELISA, but did not find detectable IFNα levels. However, T4 and TSH levels demonstrated severe primary hypothyroidism in the three high copy number lines (91, 22, and 99). For example, Figure 4B shows representative hormone levels in TG+ and TG- mice from Line 91, a high expressing line. T4 levels were low in TG+ mice at the earliest time tested (~ 4 months) and were dramatically lower than in nontransgenic litter mate controls (Figure 4B). Figures 4C & 4D show markedly low T4 levels coupled with high TSH levels in Line 91 compared to an intermediate copy number line 100.
At the time of neck dissection of the high expressing lines we observed surrounding inflammatory changes coupled with small thyroid size. Fertility, morbidity and mortality were substantially improved with initiation of T4 supplemented chow that normalized T4 and TSH levels (data not shown). Histologically, thyroids from untreated hypothyroid animals at approximately 3–4 months of age showed profound inflammatory necrosis of the thyroid (Figure 4F). Figures 4E & 4H show a normal appearing thyroid, histologically and by gross inspection taken from euthyroid intermediate copy number Line 100; Figures 4F & 4I show the histology and gross appearance of the thyroid from high expressing line 91 at approximately 4 months of age without thyroid hormone replacement demonstrating the marked inflammatory necrosis of the gland. Interestingly, T4 replacement partially reversed the severe thyroid inflammation and cell death (Figures 4G & 4J) without decreasing thyroidal levels of IFNα (Figure 4K), suggesting that T4 replacement may have a thyroid protective effect in the setting of thyroid autoimmunity, as has been suggested in pregnant AITD patients (30). The partial preservation of thyroid tissue with thyroid hormone intervention in the higher expressing lines (91, 22 and 99) allowed for dissection of sufficient thyroid gland tissue for RNA isolation. Figure 4K shows the relative expression of IFNα mRNA by QPCR in 5 of the 6 lines. Lines 91 and 22, requiring thyroid hormone treatment for survival and fertility, demonstrated markedly high IFNα expression in a range similar to the two intermediate expressing lines 100 and 30 that were not on T4 replacement therapy at the time of sacrifice at approximately 4 months of age.
Interestingly, attempts to backcross high expressing lines, initially in a predominantly C57Bl6 background, to a CBA/J background, a strain known to be susceptible to autoimmune thyroiditis (31), were unsuccessful. Even on thyroid hormone replacement litters were infrequent, small in numbers, and most of the pups were runted and did not survive. However, intermediate expressing lines (lines 100 and 30) and a low expressing line (line 2) were successfully backcrossed up to 6 generations to CBA/J without obvious fertility or viability challenges. Taken together these data support the hypothesis that local IFNα production in a susceptible individual can induce thyroiditis and thyroid dysfunction.
To study the contribution of immune cell infiltration to the thyroidal destructive processes, representative thyroids from normal, intermediate and high expressing IFNα lines, ages 3–4 months old, were processed for H & E histochemistry as well as immunocytochemistry for specific immune cell markers. Figure 5B depicts normal appearing histology from a wild type thyroid for comparison. There were inflammatory infiltrates in tissues surrounding the thyroid glands of high expressing lines (Figure 5C) suggesting a strong thyroidal chemokine response. Examination of thyroids of high expressors showed no detectable intra-thyroidal staining for macrophages (F4/80-like receptor) or B cells (CD45R/B220). There was, however, consistent but infrequent and scattered presence of cells staining positively for the T cell marker, CD3e (Figure 5F). Hypothesizing that at the time of sacrifice most infiltrating mononuclear cells have disappeared as the thyroid was almost completely necrotic, we analyzed thyroids of day 1 old pups from high expressing lines 22 and 91 and their wild type litter mates. Figure 5D reveals marked inflammatory infiltration of the thyroid and surrounding tissues from a TG Line 91 animal; corresponding tissue section from a wild type pup is shown in Figure 5A.
To elucidate potential mechanisms by which IFNα leads to destructive thyroiditis in the TG mice, we performed a transcriptome analysis comparing the entire thyroid transcriptome in the TG mice and the wild type mice. A cohort of TG mice and a selection of non TG wild type litter mates from the high IFNα expressing line 91 without thyroid hormone replacement for ~ 2 months were sacrificed at ~8 months of age and the thyroids dissected. The small size of the TG thyroids necessitated their pooling (see Methods) prior to RNA extraction. RNA was isolated and the transcriptome analyzed by RNAseq (see Methods), a method that allows for greater sensitivity, specificity, and dynamic range than hybridization based expression microarrays, while providing the potential for characterizing the expression of all transcripts in a given RNA sample. Bioinformatic pathway analysis of the RNAseq data was performed by the Ingenuity Pathway Analysis (IPA) program (Ingenuity Systems, Redwood City, CA) and the data is available using GEO accession number GSE25115 at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE25115. As expected and corroborating the marked induction of IFNα protein in the TG thyroids, the classical IFNα signaling pathways were upregulated in the IFNα transgenic mice (p=7.08×10−8, Figure 6A). In addition to the IFNα pathway, four other pathways were also significantly upregulated: antigen presentation (p=5.13×10−6), complement system (p=5.25×10−3), Granzyme B (p=3.89×10−2), and pattern recognition receptor (p=2.04×10−3). In addition, the IL-6 pathway was upregulated but to a smaller degree, with significant upregulation of the IL-6 receptor (2.6 fold increase), albeit the expression of the IL-6 gene itself was not increased. Two pathways, arachidonic acid metabolism (p=1.95×10−4), and fatty acid metabolism (p=1.00×10−2), were downregulated. Table S1 (Online supplemental material) lists individual genes up- or down-regulated with the pathways identified by Ingenuity; Table S2 (Online supplemental material) shows the results of a subset of genes selected to confirm the validity of the RNseq results by standard QPCR. In all the genes selected for confirmation, QPCR mRNA levels matched the RNseq data (Table S2). Figure 7 depicts a representative detailed outline of one of the most significantly upregulated canonical pathway, antigen presentation.
To determine whether the transcriptome results in the TG mice reflected acute or chronic changes induced by IFNα, we determined if pathways that were affected in the TG mice might also be regulated in human thyroid follicular cells exposed to IFNα for 12 and 24 hours. Human thyroid follicular cells in primary cultures were exposed to high levels of IFNα (5,000 U/ml) for 0, 12 or 24 hours followed by RNA isolation (see Methods). The samples for the three time points were pooled at each time point and performed in duplicates (0, 12, 24 hours). Not unexpectedly, this intervention, reflecting an acute stimulation by IFNα, provoked a marked increase in classical IFNα signaling pathways as seen after the chronic IFNα exposure in the TG thyroids, (p=8.13×10−8, Figure 6B). In addition, the Ingenuity pathway analysis showed 6 other pathways that were significantly altered by IFNα (Figure 6B). Three pathways were upregulated and overlapped with the pathways increased in the TG thyroids: 1) antigen presentation (p=6.61×10−9), 2) complement system (p=1.10×10−3), and 3) pattern recognition receptors pathways (p=2.29×10−3). Two pathways were uniquely increased in the human thyroid cells exposed to IFNα: apoptosis (p=2.19×10−3), and retinoic acid receptor (p=6.17×10−2). Only one pathway, the cell-cycle pathway, was downregulated (p=5.01×10−4), most likely reflecting the decreased cell viability. Thus, the antigen presentation, complement, and pattern recognition receptor pathways were acutely activated, while other inflammatory pathways, notable Granzyme B, most likely reflected chronic exposure of the tissue to IFNα.
We selected 4 key inflammation-related genes shown to be induced chronically in both TG mouse thyroids and acutely in human thyroid cells exposed to IFNα We tested these genes by QPCR in TG mice and human thyroid cells in primary cultures. These genes were OAS1, a gene that is increased as a part of the early antiviral response induced by IFNα; TRIM21, a nuclear protein that binds to and regulates the function of IgG, and is involved with several autoimmune diseases (32–34); Granzyme B, a serine protease that is a cytotoxic T-cell marker (35–37) (Figure 8); and CXCL10, a major chemokine for the recruitment of mononuclear cells (38–42) (Figure 9). We preformed time-course and dose-response experiments (Figures 8–9); the TG mice thyroidal RNA, for this comparison, was obtained from the same RNA preparations used for surveying IFNα expression (Figure 4K). As in the RNAseq analysis, all of these 4 genes showed significant upregulation in our QPCR analysis (Figures 8–9). CXCL10 was significantly induced upon acute exposure of human thyroid cells in culture and was strikingly induced with chronic exposure of thyrocytes to IFNα in the high expressing TG lines. This suggested that CXCL10 has both acute and chronic effects on thyroidal inflammation. In contrast and as expected Granzyme B, thought to be primarily expressed by immune cells, was increased only in the TG thyroids reflecting the cytotoxic T-cell damage to thyrocytes. The increased Granzyme B levels, seen at the 6 hour time point in thyroid cell cultures, most likely represents some contamination of the culture with thyroid-resident lymphocytes, and was not observed later than 6 hours (most likely due to death of contaminating lymphocytes). Moreover, no increase in Granzyme B levels in thyrocytes was observed in our dose-response experiments (Figure 8). TRIM21, with a less marked difference among the TG lines, showed a significant increase in the cultured thyrocytes, suggesting that its effects are only acute. TRIM21 appeared to be representative of a number of this class of genes found to be induced in the RNAseq analysis. For example, in the TG mice thyroid RNA, TRIM14 and TRIM5 were increased 3.4 and 2.94 fold; respectively. In the human thyrocytes, TRIM14 was increased 2.96 fold (24 hr) and TRIM5 was increased 3.36 fold (24 hr).
The mechanisms by which IFNα induces autoimmunity are still not fully understood. In particular, the roles of immune modulation vs. target tissue effects are still controversial. Here, using thyroiditis as a model, we examined the role of endogenous induction of IFNα in an attempt to dissect the direct tissue toxic effects of IFNα vs. its immune-regulatory actions. Because interferons are induced in all cells in response to viral infection, it appears that the notion of selectively increasing interferons independent of the viral stimulus has not been frequently considered in the past (43) as a strategy to understand interferon’s pathologic mechanisms. As such, our findings have novel implications to other tissues that are exposed to high local levels of IFNα, either during infections, or as an apparent side effect of large doses of IFNα administered for chemotherapy, or when induced by the DNA damage that is associated with autoimmune diseases, such as SLE. It is likely that high local levels of IFNα in certain tissues can contribute to the development of an autoimmune response to that tissue by a bystander mechanism, as was suggested in type 1 diabetes (44).
Our results demonstrate that transgenic mice overexpressing IFNα in the thyroid develop severe inflammatory thyroiditis leading to profound hypothyroidism requiring T4 supplementation for survival. These results are consistent with a previous study showing that transgenic mice overexpressing IFNα in the pancreatic beta cells develop severe insulin-dependent diabetes (45). However, this study did not analyze the mechanisms by which IFNα induced tissue-specific autoimmunity. Interestingly, Akwa et al demonstrated that transgenic mice overexpressing IFNα in the central nervous system developed a progressive inflammatory encephalopathy associated with a predominantly lymphocytic infiltration and induction of IFNα responsive genes such as MHC class I and OAS1 (43). We found a similar inflammatory response in the thyroid accompanied by inflammatory changed in tissues adjacent to the thyroid, and, more globally, demonstrated that IFNα activated the antigen presentation, complement, and pattern recognition receptor pathways. These effects would be predicted to induce inflammatory destruction of the thyroid coupled with direct toxicity potentially through Granzyme B mediated actions.
The primary actions of IFNα in cultured thyroctyes appeared to be an early increase in the expression of thyroid-specific proteins, a significantly increased activity of several pathways involved in innate and adaptive immune responses, and, ultimately, markedly increased non-apoptotic thyroid cell death. Consistent with these in vitro findings, transgenic mice overexpressing IFNα in the thyroid displayed striking destructive changes in the thyroid that were distinctly exacerbated in an autoimmune susceptible background. Moreover, the same innate and adaptive immune response pathways were upregulated in the transgenic mice. The high local levels of IFNα in the thyroid, importantly without evidence for systemic elevation of IFNα, appear to lead also to non-immune tissue destruction, the severity of which is modulated by background autoimmune susceptibility. Our data are consistent with recent studies demonstrating that cytokines such as interferon γ can directly inhibit thyrocyte function, and suggest a new mechanism for induction of hypothyroidism in the setting intra-thyroidal cytokine secretion (46,47). Indeed, lymphocytic infiltration of the thyroid alone may not be sufficient to cause hypothyroidism (48). Even though the local levels of IFNα in the TG mice are significantly higher than the serum levels in HCV patients treated with IFNα, it is likely that the thyroidal levels of IFNα in patients are higher than serum levels due to local effect of the HCV virus itself (48).
Our in vitro studies demonstrated a consistent INFα dose-dependent increase in thyroid specific proteins, TPO, NIS, TSHR, and Tg. This induction of thyroid-antigen expression by IFNα provides an attractive potential mechanism accounting for the development of autoimmune thyroiditis (49) in the setting of thyroidal inflammation (e.g. caused by the HCV virus itself (50)). Specifically peptides resulting from the degradation of these upregulated self proteins could be presented to T-cells. In support of this possibility, a significant increase in antigen presentation pathways leading to presentation of thyroid specific peptides and activation of cytotoxic T-cells has been observed in other model systems (51). Thus, the IFNα-induced upregulation of thyroid-specific antigens and of antigen presentation pathways, coupled with IFNα induced thyroid cell necrosis observed in our tissue culture and in vivo experiments, could result in the generation and presentation of pathogenic peptides derived from thyroid specific antigens (see Figure 10). Of note, a previous study by Carracio et al showed decreased thyroid antigen expression by IFNα (52). However, in their investigation the experimental conditions were different and thyroid antigen gene expression was tested at a later time point than in our study, suggesting that the decreased expression observed most likely reflected the late thyrocyte necrotic cell death associated with IFNα exposure.
A number of cytokines were increased in the TG mice thyroids and the cultured human thyrocytes. CXCL10 levels were in particular markedly increased consistent with data showing that blood levels of this cytokine are not only in high circulating concentrations in autoimmune thyroiditis (53), but also correlate with a greater probability of thyroid dysfunction in IFNα treated HVC patients (39). More recently, Antonelli, similar to our studies, has shown that IFNα markedly induces CXCL10 in cultured human thyrocytes (54). CXCL10 is associated with a number of conditions characterized by the recruitment of cytotoxic lymphocytes such as cutaneous lupus erythematosis (55), and regression of melanoma (56). Another relevant cytokine pathway that was increased in the TG mice was the interleukin 6 (IL-6) pathway, a pathway that has been shown to play a role in autoimmune thyroiditis (16,57).
Of particular interest was the finding of a marked induction by IFNα of the Granzyme B pathway in the TG thyroids. Granzyme B is a cytotoxic T-cell marker, and is a part of a well-established pathway utilized by killer lymphocytes to destroy target cells (58). Typically, Granzyme B is thought to be injected into target cells from recruited cytotoxic T-cells via perforin channels formed on the cell membrane. Apoptosis rapidly ensues (59). Moreover, recent data suggest that Granzyme B can degrade target cell proteins; as a result pathogenic peptides released during the degradation can augment the autoimmune response (60). This mechanism of autoimmune tissue destruction has been confirmed to operate in several well-characterized autoimmune diseases in which pathogenic peptides are generated, including beta cell destruction in type 1 diabetes (T1D) (61), flares of SLE (62), and multiple sclerosis (63). Our results suggest that in vivo exposure of the thyroid to increased IFNα, as a result of infection or interferon therapy, may induce locally high levels of Granzyme B derived from cytotoxic T-cells. The high levels of Granzyme B may generate pathogenic thyroidal peptides that can trigger and/or augment a destructive autoimmune response.
In the TG mice and cultured human thyrocytes exposed to IFNα, we observed cell death without evidence for apoptosis or autophagy and, therefore, by exclusion, it is likely to be by necrosis. Cell death can occur by at least three primary mechanisms: apoptosis, autophagy, and necrosis (64). Apoptosis is a programmed cell death triggered by specific receptors and signaling pathways (65), whereas autophagy is a process of sequestration of cytoplasmic constituents including oganelles which leads to their eventual degradation in lysosomes and cell death (27). Death by necrosis does not have a uniform definition and is often identified as the cause of cell death by excluding apoptosis and autophagy. While the dogma has been that necrosis is a non-programmed uncontrolled relatively end stage form of cell death, there is growing evidence for the regulation of cell necrosis by discrete signaling pathways (66). In the TG mice the thyrocyte necrosis might have been induced by cytotoxic T-cells; however, our findings in thyrocytes in culture suggested that IFNα also induced thyroid cell death by direct toxicity.
The downregulation of the arachidonic acid and fatty acid metabolism pathways is not as clearly linked to thyroid inflammation. This was observed dramatically but only in the TG thyroid tissue. Whether, these changes are epiphenomena or directly involved in the thyroid pathology is not clear but could represent decreased synthesis of membranes in the cells affected by the inflammatory damage.
Interestingly, thyroid tissue destruction increased when the high expressing transgenic (TG) lines were back-crossed from a predominantly C57BL/6 background to an autoimmunity susceptible CBA/J background. This thyroid destruction, possibly coupled with other harmful inflammatory changes in surrounding tissues, prevented the establishment of high expressing lines on the CBA/J susceptible background and suggested a strong genetic susceptibility to the thyroid toxic effects of IFNα. This notion is supported by the observations that the presence of TAb’s, considered a preclinical marker of genetic susceptibility to AITD (67), prior to the initiation of IFNα therapy, is a significant risk factor for the development of AITD during IFNα treatment (20). Moreover, we have shown that injecting IFNα to NOD H2h4 mice, a strain genetically susceptible to spontaneous autoimmune thyroiditis, caused a higher frequency of autoimmune thyroiditis, supporting the notion that IFNα triggers thyroiditis in genetically susceptible individual (68). Finally, one study by our group showed an association between several immune-regulatory genes and IIT (12).
Our results (Supplemental Table I) demonstrated a broad induction of the TRIM (tripartite motif containing) family of proteins. There are over 64 members of the TRIM family in mice (69), and, though primarily described in the context of the antiviral responses of interferons (70), their functions have been increasingly linked to an extensive range of biological actions including innate immunity. A subset consisting of 5 members of the TRIM family was found to be increased by RNAseq in the TG thyroids. Of particular interest was TRIM21 (also known as Ro52 (71)), a nuclear protein that binds to and regulates the function of IgG, and is a target autoantigen in several autoimmune diseases, most notably Sjogren’s syndrome and SLE. Moreover, genetic polymorphisms of TRIM21/Ro52 have been associated with the onset of Sjogren’s (72) and SLE (73). More recently, TRIM/Ro52−/− mice were shown to develop evidence for SLE that could be reversed by disrupting the IL23/IL17 pathway, thought to mediate some of the inflammatory changes in SLE (74). Our studies showed a dose and sustained time-dependent increase of TRM21 in response to INFα in cultured human thyroctyes and a greater expression in the higher INFα expressing TG lines (Figure 8).
Taken together, our data suggest that IFNα induces tissue specific autoimmunity by direct tissue toxic effects as well as by immune recruitment bystander mechanisms. The most notable direct thyroid tissue effect in our IIT model was the induction of thyroid cell necrosis as early as day one of age, coupled with increased expression of thyroid specific proteins. The immune effects of IFNα triggered the recruitment of inflammatory cells presumably by induction of various cytokines and chemokines, such as CXCL10, and activation of cytotoxic T-cells, causing thyroid cell necrosis at least in part through the activation of the Granzyme B pathway (Figure 10). In addition IFNα treatment, likely in combination with HCV infection (75) causes early upregulation of antigen presentation pathways, cytokines/chemokines, and cytotoxic T-cells, as well as thyroid specific antigens (e.g. TSHR, Tg), followed by late thyroid cell death by necrosis. Genetic susceptibility determines whether the ultimate response is autoimmune thyroiditis, destructive necrosis, or resistance. In genetically susceptible individuals, these IFNα-mediated effects trigger autoimmune thyroiditis (76). The association of INFα induced genes with both an appropriate anti-viral response and with pathologic autoimmune phenomena demonstrates, as suggested by Akwa (43), that IFNα is a “two-edged sword” that provides protection from viral illness and yet can induce tissue injury. Further study of our model of overexpression of INFα in the thyroid, may help distinguish the beneficial vs. destructive components of interferon actions.
We thank Dr. Robert Franco and Mary Palascak (University of Cincinnati) for assisting us with the FACS analyses, Rita Angel (University of Cincinnati) for performing the H&E and immunohistochemistry staining, and Drs. Roy Weiss and Samuel Refetoff (University of Chicago) for performing the TSH assays.
This work was supported in part by grants DK61659, and DK073681 from NIDDK, and by a VA Merit award (to YT).