Note: The FDA-cleared TSI-Mc4 bioassay is now commercially available (Thyretain™).
Professor G. J. Kahaly, Gutenberg University Medical Center, Langenbeckstreet 1, 55101 Mainz, Germany. E-mail: email@example.com
Thyroid-stimulating immunoglobulins (TSI) are a functional biomarker of Graves' disease (GD). To develop a novel TSI bioassay, a cell line (MC4-CHO-Luc) was bio-engineered to constitutively express a chimeric TSH receptor (TSHR) and constructed with a cyclic adenosine monophosphate (cAMP)-dependent luciferase reporter gene that enables TSI quantification. Data presented as percentage of specimen-to-reference ratio (SRR%) were obtained from 271 patients with various autoimmune and thyroid diseases and 180 controls. Sensitivity of 96% and specificity of 99% for untreated GD were attained by receiver operating characteristic analysis, area under the curve 0·989, 95% confidence interval 0·969–0·999, P = 0·0001. Precision testing of manufactured reagents of high, medium, low and negative SRR% gave a percentage of coefficient-of-variation of 11·5%, 12·8%, 14·5% and 15·7%, respectively. There was no observed interference by haemoglobin, lipids and bilirubin and no non-specific stimulation by various hormones at and above physiological concentrations. TSI levels from GD patients without (SRR% 406 ± 134, mean ± standard deviation) or under anti-thyroid treatment (173 ± 147) were higher (P < 0·0001) compared with TSI levels of patients with Hashimoto's thyroiditis (51 ± 37), autoimmune diseases without GD (24 ± 10), thyroid nodules (30 ± 26) and controls (35 ± 18). The bioassay showed greater sensitivity when compared with anti-TSHR binding assays. In conclusion, the TSI-Mc4 bioassay measures the functional biomarker accurately in GD with a standardized protocol and could improve substantially the diagnosis of autoimmune diseases involving TSHR autoantibodies.
The aetiology of autoimmune thyroid disease is a story of genes and environment . Thyroid-stimulating immunoglobulins (TSI), functional biomarkers in this interplay of anti-self reactivity [2,3], were described originally as long-acting thyroid stimulators (LATS) [4,5]. Later, TSI were found to stimulate thyrocytes via the thyroid-stimulating hormone TSH receptor (TSHR) either on a rat thyrocyte cell line [6,7] or on primary human thyrocytes . Stimulation was determined by measurement of radioactive cyclic adenosine monophosphate (cAMP) production [6–8] or by quantification of the metaphase index . In subsequent studies transgenic animal cells were designed to express human TSHR and to quantify the stimulatory capacity of TSH and TSHR autoantibodies [9–12]. The evaluation of TSI bioassays with Graves' disease (GD) and control sera confirmed the clinical relevance of stimulating and blocking immunoglobulins [13–19]. Major advance in the bioassay came from optimized numbers of human TSHR and the use of cells stably transfected with a cAMP-dependent luciferase reporter gene [18,19]. Most of these bioassays require incubation in serum-depleted growth medium prior to addition of patient serum and take several days to perform. Furthermore, the cAMP response of wild-type TSHR is susceptible to immunoglobulins having blocking activity .
The TSI-Mc4 bioassay in this report is based on the performance of previous chimeric constructs of human TSHR . One particular chimeric receptor, denoted Mc4, had amino acids 261–370 within the N-terminal extracellular domain of TSHR replaced by sequences from the rat luteinizing and chorionic gonadotrophin hormone receptor. TSI induced similar levels of cAMP in cells stably transfected with Mc4 TSHR and the wild-type TSHR. However, the level of TSH-induced cAMP production was five- to 10-fold lower in Mc4 TSHR cells and there was a conspicuous loss of responsiveness to the inhibitory immunoglobulins present in the serum of patients with immune thyroiditis . We engineered a stable cell line (Mc4-CHO-Luc) that contains a derivative of the Mc4 TSHR gene under a constitutive promoter and a luciferase reporter gene under control of a cAMP-dependent promoter. This cell line was then configured in a format of cryopreserved cells that after thawing and plating rapidly regenerates viable monolayers suitable for TSI measurement. In this report we assess the analytical and clinical performance of this TSI-Mc4 bioassay.
Materials and methods
Chimeric human TSHR luciferase reporter gene constructs
The neomycin resistance gene for antibiotic selection was isolated from vector pMC 1 (Stratagene, Cedar Creek, TX, USA) and subcloned into plasmid pSG5-MC4 that contains the chimeric TSHR/LHGHR driven by the SV40 promoter to generate pMC4-neo . The human glycoprotein hormone (GPH) alpha subunit promoter was amplified by polymerase chain reaction (PCR) using human embryonic kidney (HEK) cell chromosomal DNA as a template. The 316 base pairs (bp) amplicon was cloned into pcDNA2·1 vector (Invitrogen, Carlsbad, CA, USA) and inserted upstream of the firefly luciferase gene (Luc) in vector pGL2 (Promega, Madison, WI, USA). Finally, a fragment containing the GPH promoter-Luc DNA fragment was subcloned into the plasmid pMC4-Neo to generate the final plasmid, pMC4-GHP-Luc (Fig. 1). The nucleotide sequences of both the GPH promoter and the chimeric Mc4 TSHR gene were determined (Davis Sequencing Inc. Davis, CA, USA) and compared with the National Center for Biotechnology Information (NCBI) database.
Reporter cell line
Chinese hamster ovary (CHO)-K1 cells [American Type Culture Collection (ATCC) number: CCL-61; Manassas, VA, USA] were transfected with linearized pMc4-GPH/Luc plasmid using HyFect (Denville Scientific, Metuchen, NJ, USA), according to the manufacturer's instructions. Twenty-four hours after the transfection the cells were placed under antibiotic selection with 1·0 mg/ml geneticin (Sigma, St Louis, MO, USA). Clones were screened for luminescence relative light units (RLU) in triplicate wells with either control serum or serum from GD known to contain TSI. Two clones were selected that exhibited the highest luminescence index (LI), as defined by the following formula:
Following two rounds of limiting dilution cloning, clones were re-isolated and tested further for their stability upon successive passages and freeze–thaw cycles. A single clone, Mc4-CHO-Luc, that had the highest average signal-to-background ratio, was chosen for the development of the TSI-Mc4 bioassay. All subsequent experimental values were obtained with the final selected clone. Average of triplicates were normalized to the blank and the percentage specimen-to-reference ratio (SRR%) was calculated with the formula: SRR% = average TSI specimen RLU/average reference standard RLU × 100.
TSI-Mc4 reporter cell bioassay
The TSI-Mc4 bioassay was performed in the inner 48 wells of a 96-well plate with black-wall, clear-bottomed wells. Twelve serum samples and four controls, consisting of reference standard bovine TSH (bTSH), normal serum, positive TSI serum and cells alone, were tested in triplicate. The reference standard of bTSH (between 0·0156 and 0·0625 IU/l) showed stimulating activity within the range of a recombinant human TSH (rhTSH) standard curve [World Health Organization standard, National Institute for Biological Standards and Control (NIBSC) code 03/192 for bioassay]. The upper and lower limits of the range were the RLU value measured from cells induced with 0·2 IU rhTSH/l and the RLU measured from cells induced with normal serum, respectively. Normal serum, or a pool of up to five normal sera, pre-screened for SRR% < 80% of the bTSH reference standard, was used by the manufacturer as the normal control. A single donor serum that had been diluted to the targeted range of SRR% 247–605 was used as the TSI positive control.
Before planting the cells, the wells were treated with cell attachment solution (CAS; DHI, Athens, OH, USA). The frozen cells were thawed and seeded immediately at 6·7 × 104 cells/well and then incubated in growth medium supplemented with 40 µM dexamethasone to reduce the background cAMP signalling (GM, DHI) at 37°C with 5% CO2 for 15–18 h before addition of the samples. Positive control serum, normal control serum, the index patient test sera and the bTSH reference control were prepared by adding one part sample to 10 parts optimized reaction buffer containing 8% polyethylene glycol, molecular weight (MW) 8000.
After 3 h, the luciferase expression levels of cell lysates were measured directly in the wells following addition of substrate and lysis reagent (Promega, Madison, WI, USA) using a multi-well plate luminometer (Veritas Microplate Luminometer, Turner BioSystems, Sunnyvale, CA, USA or Tecan Infinite M200, Tecan GmbH, Crailsheim, Germany). After automated shaking of the plates for 5 s, the luminescent measurements were made over 1 s integration time at 22–24°C. All samples were tested in a blinded manner at two independent testing sites. The laboratory technicians received a training course and passed a quality control evaluation in which they achieved coefficients of variation (CVs) < 8% for all triplicate measurements of 26 samples run on 2 consecutive days.
Interference and cross-reactivity
For the interference studies, the TSI detecting cells were induced with bilirubin (Sigma 055K0919), serum haemoglobin (Sigma 037K7675), prednisolone (Merck batch no. 7569401) or lipaemic human serum in low or high levels of TSI-containing sera. For the cross-reactivity studies, the MC4-CHO-Luc cells were treated with luteinizing hormone (LH), human chorionic gonadatrophin (HCG), follicle-stimulating hormone (FSH) and recombinant human thyroid-stimulating hormone (rhTSH) in high or low levels of TSI containing sera, respectively. All the hormones were from NIBSC.
TSH receptor autoantibodies (TRAb) assays
The anti-TSHR binding in sera were measured by coated tube kit (Kronus, Boise, ID, USA). TRAb positivity was reported for values greater than 1·1 IU/l. Sera tested with the TRAK RIA that gave discordant results with the TSI-M4 bioassay or concordant sera selected for the comparison of TSI and TRAb titres by dilution analysis were retested with the automated Cobas electrochemiluminescence ECLIA Elecsys, an anti-TSHR immunoassay that uses a porcine TSHR and human anti-TSHR autoantibody M22 (Roche Diagnostics GmbH, Penzberg, Germany). TRAb positivity was greater than 1·5 IU/l [20,21].
Sampling of sera from consecutive patients with clinically and biochemically confirmed GD (n = 96, 55 female, mean age 44 years, range 13–75 years), Hashimoto's thyroiditis (HT, n = 62, 39 female, 47 years, range 16–74), systemic lupus erythematosus (SLE, n = 17, 12 female, 36 years, range 24–50), rheumatoid arthritis (RA, n = 13, 12 female, 69 years, range 67–70), type 1 diabetes (T1D, n = 36, 12 female, 33 years, range 12–69), chronic type A autoimmune gastritis (CAG, n = 19, 14 female, 52 years, range 14–71), thyroid nodules (TN, n = 36, 24 female, 40 years, range 18–61) and control sera of healthy euthyroid blood donors (n = 180, 94 female, 25 years, range 3–68) were obtained with signed informed consent. Blood sampling was approved by the local State Ethical Committee. All sera were stored in aliquots at −20°C until measurement.
Dilution of sera and analysis of anti-TSHR autoantibodies
The titres of TRAb and TSI in selected patient serum were determined by making serial dilutions of the patient serum into normal control serum and the TSI determined by the bioassay after addition of one part of the neat serum or diluted serum into 10 parts reaction buffer, as described above. The TRAb were measured directly by ECLIA Elecys.
All the data were analysed by either template software (Veritas Microplate Luminometer Software, version 1·7·1) or the Tecan instrument control and data analysis software (Magellan Tracker, version 2·4). The TSI specimen was the cells induced with diluted serum samples (1:11); the reference RLU was the cells induced with bTSH at 0·031 mIU. Above normal SRR% was determined to be ≥ 140% above the reference. For each test, the percentage CV (CV%) was calculated according to the formula:
The sensitivity and specificity of the assay was obtained by receiver operator curve (ROC) analysis using the web-based MedCalc software version 11·1. Comparisons of the TSI values between patient groups were assessed by Student's t-test. A two-sided P-value < 0·01 was considered to be statistically significant with values of each group or differences between groups reported in 95% confidence intervals.
The limit of detection (LoD) for the assay encompassing 95% [2 standard deviations (s.d.)] of all values was 89·4 SRR% according to the formula LoD = LoB SRR% + 1·645 × s.d.low, where by limit of the blank (LoB), the average of 20 blank measurements, was SRR% 62·75 and s.d. low; s.d. of 25 repeated measurements of the low positive TSI serum specimen was SRR% 16.
Verification of the Mc4-CHO-Luc final clone
The serum of GD patients adsorbed with Protein G Sepharose did not trigger cAMP above the background level in the Mc4-CHO-Luc cells indicating that the cAMP produced by this clone is attributable to the immunoglobulin G fraction in serum (results not shown). The cell line exhibited a dose–response diminution in the level of luciferase activity after serial dilution of TSI-containing sera (Fig. 2a). The Mc4-CHO-Luc cell line was stable after successive passages in cell culture without reduction in the level of TSI-inducible luciferase activity after each passage up to 20 passages in the absence of antibiotic selection (Fig. 2b). The luciferase activity measured in the CHO-Luc cell line containing the luciferase reporter gene without transfection with the Mc4 TSHR shows no increase in RLU upon additions of either bTSH or serum of GD patient (Table 1).
Table 1. Luciferase activity of CHO Luc cells without thyroid-stimulating hormone receptor (TSHR) construct.
Control luciferase activity (RLU)
The mean, standard deviation (s.d.) and range of triplicate measurements of the luciferase activity in relative light units (RLU), were obtained in the Chinese hamster ovary (CHO) Luc cell line transfected with the luciferase reporter gene without the Mc4 TSHR construct . Sample additions of bovine TSH or serum of Graves' disease (GD) patient (GD 308) were made after dilution into reaction buffer as described in the Materials and methods section.
bTSH 33 mU/l
bTSH 100 mU/l
bTSH 1000 mU/l
Intra- and interassay precision
Intra-assay precision of the TSI-Mc4 bioassay, test results obtained by the same user within the same day repeated in two plates, gave a CV of 5% with normal serum (SRR% 61), 4·2% with low-positive TSI serum (SRR% 183), 2·6% with mid-range TSI serum (SRR% 313) and 3·6% with high TSI serum (SRR% 509). To determine the interassay precision, the same set of samples was tested repeatedly across a 20-day period and the data was collected by the same user. This testing gave an overall average interassay CV of 12% (Table 2). The interassay precision was evaluated for its reproducibility. Positive serum, bTSH reference and normal serum were tested in 44 plates at two sites by three users over a period of 6 months. CVs of 4·4%, 9·3% and 7·1%, respectively, were obtained.
Table 2. Interassay precision.
Thyroid-stimulating immunoglobulin (TSI)-containing sera of Graves' disease patients were diluted with negative control serum (negative) to yield percentage of specimen-to-reference ratio (SRR%) values near the cut-off (low), slightly higher that the cut-off (medium) and near the upper level of the TSI-Mc4 bioassay detection (high). Each sample was tested repeatedly 120 times by a single user over a period of 20 days. CV, coefficient of variation; s.d., standard deviation.
Specificity, cross-reactivity and interference
G-protein-coupled glycoprotein hormones that share identical alpha chain structure with TSH were tested for interference and cross-reactivity in the TSI-Mc4 bioassay. TSI-positive or negative serum samples were spiked with increasing concentrations of the purified pituitary hormones LH, FSH and rhTSH or with the reproductive hormone HCG. No cross-reactivity or interference was found with these hormones, even at concentrations of 1–2 log units above the physiological range (Fig. 3a–c). Interference was seen only at very high concentrations of rhTSH (Fig. 3d). To determine whether various substances found in serum might interfere with the bioassay, sera containing positive or negative TSI were spiked with increasing concentrations of haemoglobin, bilirubin, lipids or prednisolone (Fig. 3e–h). None of these substances had a significant effect on the SRR% when present within the normal range or at concentrations expected in patient serum. However, a decline in the TSI activity was caused by supraphysiological concentrations of lipids (> 3000 mg/l), haemoglobin (> 2500 mg/l) and bilirubin (> 365 mg/l). Prednisolone had no inhibitory effect.
ROC analysis and determination of the TSI cut-off
The sera of 54 patients with untreated GD and the sera of 180 control donors comprised the ‘training set’ for the ROC analysis which gave an area under the curve of 0·989, 95% CI 0·969–0·999, P = 0·0001 (Fig. 4). All control sera had SRR% of less than 120 and 52 of 54 GD sera gave SRR% > 150. Thus, any serum tested with the TSI-Mc4 bioassay was considered positive for the presence of TSI if the resultant SRR% measured greater than or equal to 140% of the reference control bovine TSH, a value that corresponds to > 3 s.d. above the mean of control serum.
Distribution of TSI levels
The clinical sensitivity and specificity of the TSI-Mc4 bioassay were determined by measuring SRR% values of various patient groups relative to the cut-off of 140 (Fig. 5). Fifty-two of 54 patients with untreated GD tested TSI-positive, yielding a clinical sensitivity of 96%. All 180 sera from healthy controls (100%), 85 of 85 patients with autoimmune diseases without thyroid disorders and 36 of 36 patients with thyroid nodules tested negative. In addition, 61 of 62 sera from HT patients (98%) tested TSI negative. The TSI levels, SRR% (mean, range), of the GD patients without (414, 34–660) and with (141, 78–487) anti-thyroid treatment were markedly higher than those with HT (47, 21–119), SLE (26, 19–36), RA (44, 24–69), T1D (20, 8–32), CAG (20, 15–27), thyroid nodules (54, 35–82) and controls (35, 5–116) (P < 0·0001, Fig. 5). Also, TSI levels in untreated GD were higher than those in treated disease (P < 0·001, 95% CI 176–190, Fig. 5).
Comparison of the TSI-Mc4-bioassay with TRAb assays
The bioassay was compared to a radioreceptor binding assay (TRAb; Kronus). The positive percentage agreement was 87% (95% CI, range 88·2–96·8%). The negative percentage agreement was 94% (84·0–93·2%) with 48 of 558 (8·6%) and 28 of 558 (5%) of the sera giving discordant or indeterminate test results, respectively (results not shown).
The analytical sensitivity of the TSI-Mc4 bioassay was compared with three TSHR-binding assays (Fig. 6). Sera from well-defined GD patients were diluted serially into normal serum and tested in the bioassay and in TSHR-binding assays. Two of the sera that had a positive concordance between the two methods (TSI-positive/TRAb-positive) showed an extinction of their binding activity between the dilution 1:11 and 1:22, but were still positive in the bioassay at a dilution of 1:300 (Fig. 6a). Two discordant sera (TSI-positive/TRAb-negative) both had a TSI titre > 300, but differed in the pattern of their dilution curves (Fig. 6b).
In the present work, we validate a newly engineered cell-based TSI bioassay. The derivative chimeric TSHR in this luminescent reporter bioassay, amenable to automation from a frozen and ready-to-use format, allows the clinical laboratory to synchronize the plating of Mc4CHO cells and 16 h overnight culture with the performance of the bioassay on the same day that the patient's serum samples are collected. The bioassay's standardized protocol is more efficient compared with cumbersome radioactive end-point and the labour-intensive cell culture of previous TSHR bioassays [5–10]. This novel bioassay demonstrates a high degree of intra- and interassay precision. The reproducibility, assessed by triplicate measurements made by multiple users from two independent testing facilities, gives a CV of less than 10%. This low CV indicates that the TSI-Mc4 bioassay and its components perform consistently in different laboratories and with different operators.
The clinical sensitivity of this TSI-Mc4 bioassay is consistent with previous TSI bioassays and anti-TSHR binding assays that typically show sensitivity of 95–98% in untreated GD [20,21] and TSI positivity of 48–80% among patients under treatment with anti-thyroid medications [12,18–20,23]. The striking absence of TSI activity in 61 of 62 (98%) HT patients, 301 of 301 (100%) healthy donors and patients with non-endocrine autoimmune conditions underscore the bioassay's high diagnostic specificity.
The diagnostic accuracy of TSI-Mc4 bioassay was compared previously with the reference method that employs a fully intact wild-type human TSHR (wtCHO) . Although the analytical sensitivity of the two bioassays was similar, the TSI-Mc4 bioassay exhibited significantly greater clinical sensitivity than the wild-type bioassay for the detection of TSI among patients with severe GD . In the present work, analytical sensitivity was evaluated further using sera of patients with discordant (TSI-positive/TRAb-negative) or concordant (TSI-positive/TRAb-positive) results. Sera were diluted into normal serum and the levels of TSI were compared with the levels of anti-TSHR binding activity. TSI titres were 10–30-fold higher than TRAb binding titres, indicating a greater sensitivity of the TSI-Mc4 bioassay compared with the TRAb assay in GD patients.
High intermethod variation has been reported among the different assays of anti-TSHR binding, especially in patients in relapse or remission . TRAb methods measure the ability of immunoglobulins in patient serum to compete with the binding of either radio-labelled TSH [17,25] or a luminescent-labelled monoclonal antibody [20,21,25] and they provide, at best, indirect evidence of the disease-specific autoantibodies in GD. In a few cases, displacement of ligand binding has little or no relevance to the autoimmune process ongoing in patients during anti-thyroid treatment. TRAb values can be similar in GD patients, whether in remission or during relapse . In contrast, the TSI-Mc4 bioassay distinguishes between these two groups with high levels being associated with relapse and negative or borderline levels seen among patients in remission or under immunosuppressive therapy .
There was a statistically significant difference between the TSI levels of untreated GD patients compared with the levels of patients undergoing treatment. The latter are distributed in two distinct clusters, with ranges of SRR% 30–130 and 280–490 below and above the cut-off, respectively. These differences in TSI are probably attributable to two groups of patients: (i) a group with persistently high TSI levels associated with extrathyroidal manifestations; and (ii) a group with negative TSI or low positive TSI levels that represents GD patients without systemic involvement who have been rendered either euthyroid or who are in remission . Although TSI levels correlate with pathogenesis of GD [3,26–29], the clinical relevance of determining TSI titres in GD requires further clarification. To determine whether or not TSI levels and TSI titres are predictive of disease progression, remission or response to therapy, a prospective study to monitor TSI at the onset of GD and during treatment follow-up is ongoing at our institution.
When compared to TRAb, TSI levels correlated more effectively with personalized risk factors of GD, e.g. smoking, severity of thyrotoxicosis and associated autoimmune diseases . A definite link between thyroid and orbital autoimmunity is pending; however, it is likely that circulating TSI bind to TSHR-expressing target cells in various organs leading to activation and hypertrophy . Activated TSHR-expressing target cells subsequently release hydrophilic acid muccopolysaccharides leading to oedema and swelling of the target tissue. Hence, TSI seems to be a sensitive biomarker for severity and/or systemic involvement of GD, possibly by enhancing the release of pathogenic cytokines and tissue destructive cellular infiltration driven by dysregulated antigen-specific autoagressive T cells [27–29].
In conclusion, this novel TSI-Mc4 bioassay standardizes the measurement of TSI and could improve substantially the diagnosis of autoimmune diseases involving TSHR autoantibodies.
We are grateful to L. Grippa, J. Houtz, A. Larrimer and T. Curtiss, Diagnostic Hybrids Inc. (DHI), Athens, OH, USA, as well as to M. Kanitz, Gutenberg University Medical Center, Mainz, Germany for valuable technical assistance.
S. D. L., L. D. K. and G. J. K. consult for DHI. Y. L. and P. D. O. are DHI employees.