Hyperthyroidism, defined by overproduction of thyroid hormones, has a 2–3% prevalence in the population. The most common form of hyperthyroidism is Graves' disease. A diagnostic biomarker for Graves' disease is the presence of immunoglobulins which bind to, and stimulate, the thyroid stimulating hormone receptor (TSHR), a G-protein coupled receptor (GPCR). We hypothesized that the ectopically expressed TSHR gene in a thyroid stimulating immunoglobulin (TSI) assay could be engineered to increase the accumulation of the GPCR pathway second messenger, cyclic AMP (cAMP), the molecule measured in the assay as a marker for pathway activation. An ectopically expressing TSHR-mutant guanine nucleotide-binding protein, (GNAS) Chinese hamster ovary (CHO) cell clone was constructed using standard molecular biology techniques. After incubation of the new clone with sera containing various levels of TSI, GPCR pathway activation was then quantified by measuring cAMP accumulation in the clone. The clone, together with a NaCl-free cell assay buffer containing 5% polyethylene glycol (PEG)6000, was tested against 56 Graves' patients, 27 toxic thyroid nodule patients and 119 normal patients. Using receiver operating characteristic analysis, when comparing normal with Graves' sera, the assay yielded a sensitivity of 93%, a specificity of 99% and an efficiency of 98%. Total complex precision (within-run, across runs and across days), presented as a percentage coefficient of variation, was found to be 7·8, 8·7 and 7·6% for low, medium and high TSI responding serum, respectively. We conclude that the performance of the new TSI assay provides sensitive detection of TSI, allowing for accurate, early detection of Graves' disease.
Autoimmune thyroid diseases (AITDs) are the most common of all autoimmune diseases . AITD traditionally includes autoimmune hyperthyroidism (Graves' disease) and autoimmune thyroiditis, or Hashimoto's disease (HT) . Autoantibodies to three principal thyroid antigens, thyroglobulin (Tg), thyroid peroxidase (TPO) and thyroid stimulating hormone receptor (TSHR), are found in AITD patients , as well as 18% of the asymptomatic normal population . Autoantibodies to the TSHR (TSH receptor antibody: TRAb) may be stimulating, blocking or neutral . Stimulating antibodies mimic the action of thyroid stimulating hormone (TSH) and cause hyperthyroidism (Graves' disease), whereas blocking antibodies block the binding of TSH and cause hypothyroidism [5–7]. Both stimulating and blocking antibodies can occur together in any given patient and may change over time [3,8,9].
Several different assays are used to measure TRAb , including functional cell-based bioassays . In these assays, cells expressing the TSHR on their plasma membrane are incubated with test serum and any resultant stimulation of the G-protein coupled receptor (GPCR) pathway is measured by directly measuring the pathway second messenger, cyclic AMP (cAMP), or indirectly using a cAMP responsive promoter element driving a reporter gene. These assays are capable of differentiating stimulatory from blocking and neutral antibodies and their use in clinical laboratory practice has the potential to predict relapsed patients or Graves' disease remission more accurately .
In this report, we looked to improve upon the efficiency of an early version of the cell-based bioassay . We hypothesized that some genetic alterations to the TSHR expression cassette would aid accumulation of cAMP, and thus sensitivity. A new Chinese hamster ovary (CHO)-based cell clone was engineered to express ectopically a copy of human TSHR fused to its respective GalphaS subunit with a GTPase-inactivating mutation. This new clone, together with changes to the cell assay buffer, produced marked improvements in the efficiency of the thyroid stimulating immunoglobulin (TSI) assay.
Materials and methods
Low, medium and high TSI response sera were pooled from patient samples after production runs with the existing TSI assay. Serum pooled for low, medium and high response had results between 130–150%, 151–300% and greater than 300% of normal, respectively.
Pooled normal human sera was obtained from Millipore (Millipore, Billerica, MA, USA; cat. no. S1-100 ml, lot no. NG1820782).
Normal donor sera were obtained from ARUP blood services (ARUP Laboratories, Salt Lake City, UT, USA). Hyperthyroid sera were collected from patient sera after scheduled testing had been completed. Sera were considered for the hyperthyroid–Graves' population if TSH < 0·3 mU/l, free T4 > 1·7 ng/dl and thyroid stimulating hormone TRAb > 1·75 IU/l. Sera were considered for the toxic thyroid nodule population if TSH < 0·3 mU/l, free T4 > 1·7 ng/dl and thyroid stimulating hormone receptor antibody (TRAb) ≤ 1·75 IU/l. Treatment status for the hyperthyroid patients was unknown.
The base plasmid, pVT1760, is a Moloney murine leukaemia virus (MMLV)-based plasmid with puromycin resistance and transgene expression driven by a copy of the HIV2 tat promoter . Copies of the full-length TSHR and guanine nucleotide-binding protein, (GNAS)-L cDNAs were purchased from Origene Technologies (Rockville, MD, USA). The cDNA inserts were amplified by polymerase chain reaction (PCR) with primers which added restriction sites to the amplicon ends outside the coding region to expedite cloning. One additional PCR reaction, performed while amplifying the GNAS cDNA, created a R201H mutation by overlap extension. Upon insertion of both TSHR and either GNAS PCR fragments into pVT1760 by standard ligation technique, the resultant plasmid insert was sequenced to verify the expected sequence.
Creation of TSHR/GNAS-expressing clones
Pantropic retrovirus was made by combining either pVT1760–TSHR/GNAS or pVT1760–TSHR/GNAS R201H plasmid (6·25 µg) with pVSVg (18·75 µg) and transfected into 70% confluent gp2-293 packaging cells (Clontech Laboratories, Mountain View, CA, USA) in a 10-cm dish using Transit 293 (Mirus Bio, Madison, WI, USA), following the manufacturer's protocol. Two days post-transfection, the supernatant was harvested and passed through a 0·45 µ cellulose acetate syringe filter. The subsequent filtrate was frozen at −80°C until used.
CHO1542 cells, CHO cells ectopically expressing HIV2 tat protein, were used for transduction. A mixture of 30% viral supernatant and 6 µg/ml polybrene were incubated on a 50% confluent culture for 24 h, followed by a media change with regular growth media containing 7·5 µg/ml puromycin. The puromycin selection continued for 2 weeks, during which time the cells were expanded when confluent.
The wild-type GNAS cells were cloned by picking puromycin-resistant colonies, expanding and testing for the best responders in the TSI assay. The resultant clone was designated CHO1·3.
The R201H GNAS cells were sorted by surface expression of TSHR using mouse anti-TSHR (Pierce Biotechnology, Rockford, IL, USA; cat. no. MA1-34582) and goat anti-mouse immunoglobulin (Ig)G Alexa 633 (Invitrogen, Grand Island, NY, USA; cat. no. A21053) on a FACSAria II (BD Biosciences, San Jose, CA, USA). A more detailed staining protocol is found below. The brightest staining 1·7% of the puromycin-resistant population was sorted (90% efficiency) and then cloned by limiting dilution. Individual clones were then chosen on the basis of their performance in the existing TSI assay. The clone performing best in the TSI assay was designated IIC7, and was frozen in liquid nitrogen in ready to plate aliquots.
IIC7 cells were thawed, diluted in growth media to 75 000 cells/ml, and 7500 cells/well seeded into white-sided, clear-bottomed 96-well plates (Greiner Bio-One, Monroe, NC, USA; cat. no. 655098). This number of cells yielded an optimal assay response for an overnight incubation before assay. The plates were incubated overnight at 37°C, 5% CO2. The remaining steps in the assay were carried out on a Tecan Evo 150 automated liquid handler (Tecan Group Ltd, Mannedorf, Switzerland). The following morning, the media were removed and replaced with 100 µl cell assay buffer [potassium chloride, 0·005 M; calcium chloride dihydrate, 980 µM; magnesium sulphate heptahydrate, 398 µM; dibasic sodium phosphate, 338 µM; monopotassium phosphate, 441 µM; d-glucose, 0·006 M; bovine serum albumin, 0·450% (w/v); HEPES, 0·014 M; 3-Isobutyl-1-methylxanthine (IBMX), 1 mM; polyethylene glycol MW6000, 5% (w/v)]. Test sera were diluted in cell assay buffer without IBMX, and the diluted sera then added to the wells of the cell microplate resulting in a final serum dilution of 1:18·7. The plate was incubated for 90 min at 37°C. The amount of cAMP accumulated in the cells was then measured using the HitHunter cAMP XS assay (DiscoveRx, Fremont, CA, USA), following the manufacturer's protocol.
For tests that included CHO1·3 cells, the cells were thawed, diluted in growth media to 150 000 cells/ml, and 15 000 cells/well were seeded into white-sided, clear-bottomed 96-well plates (Greiner; cat. no. 655098). This number of cells yielded an optimal assay response for an overnight incubation before assay. All other aspects of the assay were identical to the IIC7 cells.
CHO–TSHR-25 cells, in-licensed cells from Dr Leonard Kohn (Interthyr Corp., Athens, OH, USA) over-expressing the wild-type TSHR, were seeded into white-sided, clear-bottomed 96-well plates (Greiner; cat. no. 655098) fresh from cultures at 15 000 cells/well and cultured 36 h prior to assay. All other aspects of the assay were identical to the IIC7 cells.
Relative light unit (RLU) values were converted to RLU% of normal using an average from replicate values of normal serum samples included onto each plate. For precision and receiver operating characteristic (ROC) analysis, RLU% of normal data was loaded into EP Evaluator software (Data Innovations, LLC, South Burlington, VT, USA. For interference analysis, data were analysed using built-in t-test functions on Excel (Microsoft, Seattle, WA, USA). Limit of detection (LoD) followed the method of Armbruster and Pry . Briefly, limit of blank (LoB) is calculated by the equation: LoB = meanblank + 1·645 × [standard deviation (s.d.)blank]. LoD is found by the equation LoD = LoB + 1·645 × (s.d.low response sample).
Cells with GNAS fused to TSHR yield better sensitivities compared with wild-type TSHR
CHO–TSHR-25 and CHO1·3 clones (wild-type and GNAS fused TSHR, respectively) were tested in a TSI assay against 28 mixed TSI response sera. Results were calculated as RLU% of normal, and plotted on a scatterplot (Fig. 1). The responses from both assays were nearly identical up to 200% of normal, whereas above that amount the CHO1·3 cells returned higher values for the same sera than did CHO–TSHR-25 cells.
IIC7 clone and 5% polyethylene glycol (PEG)6000 cell assay buffer increase signal in the TSI assay
The IIC7 clone (R201H GNAS fused TSHR) was tested against normal serum and a low TSI response pooled serum with and without 5% PEG in the cell assay buffer. CHO cells ectopically expressing the TSHR–GNAS fusion without the R201H mutation, CHO1·3, were used as a baseline control. The results of a typical assay are shown in Fig. 2a. Using the IIC7 clone in the TSI assay resulted in a 73% increase in TSI low response serum signal, measured as RLU% of normal sera. The Z′ statistic , used as a measure of assay quality, increased from −0·33 for CHO1·3 to 0·55 for IIC7.
Inclusion of 5% PEG into the cell assay buffer that bathes the cells during the 90-min incubation with serum caused an increase in signal for both cell lines tested; 138% for CHO1·3 and 98% for IIC7. The Z′ statistic calculated for 5% PEG increased from −0·33 to 0·61 for CHO1·3 and from 0·55 to 0·86 for IIC7. More importantly, the increases realized with IIC7 and 5% PEG cell assay buffer were additive, together yielding a 243% increase over the TSI with CHO1·3 and no PEG in the cell assay buffer.
Individual sera were run on CHO1·3 TSI assay, and 28 mixed TSI response samples chosen to run on the IIC7 TSI assay with 5% PEG cell assay buffer. Resultant RLU% of normal reference values were plotted on a scatter-plot (shown in Fig. 2b). Samples in the CHO1·3 cell assay recording values less than 150% were now resulting in significantly higher values (see data points in the ellipse on Fig. 2b). Three data points, considered negative by CHO1·3 (< 111%), were now positive by the IIC7 TSI (≥ 123). No samples that were negative by IIC7 assays were positive by CHO1·3.
Simple and complex precision was checked using aliquoted normal, low, medium and high TSI response sera by making six separate runs on three different days. The cumulative results of those runs are shown in Fig. 3. Total percentage coefficient of variation (%CV) was 7·6, 8·7 and 7·8% for high, medium and low TSI response sera, respectively.
To check if the outer 36 wells on the microplate were, on average, different from the inner 60 wells, one run was made using medium TSI response sera in every position on the plate. Comparing the two populations, with the null hypothesis that the two populations are equal, a P-value of 0·16 was found. This value being greater than the significance level of 0·05 makes the null hypothesis true, and the data from inside wells not significantly different from the data from outside wells. The lack of plate positional bias allows for the entire plate to be used, aiding throughput and keeping costs down.
The calculations for LoB and LoD followed the method published by Armbruster and Pry . LoB, in this instance, is the highest apparent RLU% of normal response expected to be found when replicates of normal, healthy volunteers are tested. LoD is the lowest RLU% of normal response likely to be distinguished reliably from the LoB and at which detection is feasible. Briefly, LoB = meanblank + 1·645(s.d.blank) and LoD = LoB + 1·645(s.d.low response sera). The blank sera population consisted of average RLU% of normal values (20 replicate points per day from runs on three separate days) from normal pooled serum. For the low response population, triplicate data from four low positive sera (115–140%) from the accuracy run were used to calculate the average standard deviation. Results of the calculations showed the LoB to be 109% and the LoD to be 123%.
For an example of TSHR stimulating antibody [e.g. M22 monoclonal antibody (mAb); Kronus Inc., Boise, ID, USA], this LoD equates to between 15 and 46 pM antibody, as determined by t-test at a significance level of ≤ 0·05. Care should be taken in interpreting this result, as different antibodies have different binding affinities and may require a different amount of antibody to achieve the same result.
Clinical sensitivity and clinical specificity
Three groups of serum, normal, toxic thyroid nodule and hyperthyroid–Graves', separated on the basis of their response in TSH, free T4 and TRAB assays, were tested in TSI. The results of the ROC analysis comparing these populations are shown in Fig. 4. The normal versus hyperthyroid–Graves' ROC plot (Fig. 4a) calculated a maximum efficiency at a cut-off of 115 RLU% of normal, obtaining a sensitivity of 96·4%, specificity of 99·2%, an efficiency of 98·3% and an area under the curve (AUC) of 0·971. Toxic thyroid nodule versus hyperthyroid–Graves' ROC plot (Fig. 4b), a comparison more likely to be seen in the production laboratory, also calculated a maximum efficiency at a cut-off of 115 RLU % of normal. At this cut-off, sensitivity was 96·4%, specificity was 96·3% and efficiency was 96·4%, with an AUC of 0·966.
The calculated LoD, however, produced a cut-off of 123 RLU% of normal. At this cut-off, the statistics comparing normal versus Grave's change to 92·9% for sensitivity, while the specificity remains unchanged at 99·2%. At a cut-off of 123 RLU% of normal, statistics comparing toxic thyroid nodule versus hyperthyroid Graves' change slightly to 92·9% for sensitivity, while the specificity remains unchanged at 96·3%.
Triglycerides, bilirubin anbd haemoglobin were diluted in low, medium or high TSI response pooled sera and then assayed on the new proposed TSI bioassay as usual. Follicle stimulating hormone (FSH), luteinizing hormone (LH) and human chorionic gonadotrophin (hCG) were diluted in normal reference and medium TSI response pooled sera and then assayed on the new proposed TSI bioassay as usual. Triglycerides, bilirubin, FSH, LH and hCG did not affect significantly the response to the sera in the new TSI assay at concentrations up to 2000 mg/dl, 25 mg/dl, 200 mIU/ml, 37·2 IU/ml and 40·6 IU/ml, respectively, as determined by t-test (t-test result > 0·05; data not shown). Each of the concentrations tested of hormone agonists is above the highest upper reference interval limit (134·8 mIU/ml, 95·6 IU/l and 5 IU/l for FSH, LH and hCG, respectively).
Haemoglobin, at concentrations of 1327 mg/dl, inhibited significantly the TSI response in each of the TSI response sera into which it was diluted (t-test result ≤ 0·05). Haemoglobin was tested subsequently in a twofold serial dilution dose–response in medium TSI response serum, and was found to inhibit TSI response significantly, with haemoglobin concentrations above 332 mg/dl (Fig. 5a).
TSH, a known TSI assay interferent , was diluted in normal reference, and medium TSI response pooled sera and tested in the new proposed TSI assay in a dose–response fashion. In medium TSI response sera, TSH did not interfere significantly (data not shown). However, in normal reference sera, TSH concentrations above 24 mU/l interfered with the bioassay by increasing significantly the RLU% of normal, as determined by t-test (Fig. 5b). The CHO1·3 cells show interference with TSH serum concentrations at or above 76 mU/l, so the IIC7 with 5% PEG version of the TSI assay shows greater sensitivity to the wild-type receptor agonist.
In the present work, we validate a newly engineered cell and assay buffer for a TSI assay. Two changes to the TSHR expression cassette were investigated. First, the fusion of GNAS to a GPCR has been shown to aid accumulation of cAMP [16,17], perhaps via facilitation of coupling to the receptor and reduction of receptor recycling time. In addition, inhibition of Gαq binding to the TSHR should also aid sensitivity, as the pathway stimulated by Gαq is not measured in our detection scheme. In our experiments, cAMP accumulated within CHO1·3 cells in greater amounts in response to the same sera incubated with CHO–TSHR-25 cells, but only at higher concentrations of stimulating antibody. Previous work has suggested that a greater percentage change is obtained with cells expressing fewer receptors on the surface , due to a negative co-operativity that develops when the receptors dimerize . This might explain the observation with CHO1·3 and CHO–TSHR-25 cells; however, the IIC7 clone expresses more surface receptor than CHO1·3 (data not shown) and yet demonstrates better signal range.
The second change to the cells was to create a point mutation in the fused Gαs to inactivate the GTPase activity, thus causing prolonged stimulation of the adenyl cyclase and, therefore, increased accumulation of cAMP . While the IIC7 clone clearly outperformed the CHO1·3 clone without the R201H mutation, it would be difficult to claim that all the improvement in sensitivity resulted from the mutation. Important to the activity of the receptor is the processing (e.g. glycosylation, disulphide bond formation, cleavage, etc.) that occurs in the cell . In our hands, proper processing must have played a key role, because it became apparent during construction of IIC7 that most of the cells receiving viral construct did not express receptor on the surface (only 1·7% of the puromycin-resistant population). The insert copy number must have also played a role in proper processing, as increasing the multiplicity of infection (MOI) during viral transduction resulted in a lower percentage of the population expressing surface TSHR, despite a resultant larger puromycin-resistant population (data not shown).
The inclusion of 5% PEG6000 in the cell assay buffer played a key role in increasing sensitivity of the assay. The authors who first described the augmentative effect  never elucidated the mechanism by which inclusion of PEG stimulated sensitivity in the TSI assay. The enhancement in cAMP accumulation by PEG is vaguely reminiscent of polymer enhancement of DNA ligation . In that instance, PEG raises the effective concentration of DNA ends, and thus the likelihood of compatible ends adhering to each other. Raising the effective concentration of ligands may also be the mechanism by which the stimulating antibodies in the serum can more easily find and bind the receptors on the cells.
The previous cells used, CHO–TSHR-25, had a clinical sensitivity of approximately 80%, and a clinical specificity of 98 and 75% for normal and toxic thyroid nodule sera, respectively. The IIC7 TSI assay achieved an improved sensitivity, 93%, consistent with another published bioassay for TSI .
Consistent with an increased sensitivity to TSI was an increase in sensitivity to TSH (187 mU/l for CHO–TSHR-25 decreased to 24 mU/l with IIC7 cells). This interference level is still above the normal reference intervals for all genders and age classes (except cord blood), especially for a hyperthyroid patient (< 0·3 mU/l). In addition, the interference detected with TSH was present only when diluted in pooled normal sera, and not in sera containing TSI.
Precision improved from approximately 13% total CV with CHO–TSHR-25 to less than 9% total CV. The improved precision is probably the effect of reduced between-day error caused by a switch to using a large lot of frozen cells that are used 1 day after thawing and seeding. Indeed, the between-day error reduced from approximately 10% CV using cultured CHO–TSHR-25 cells to approximately 4% CV with the frozen IIC7 cells.
In conclusion, use of a novel TSHR–GNAS fusion-expressing cell line and modified assay buffer has achieved sensitivities in the TSI assay that should allow for fewer false negatives. The assay is completely automated and can be completed in 8 h. We are confident that this assay will improve early diagnosis of Graves' disease.
The authors would like to express their gratitude to Shelley Jenson, Tom Newell, Christopher Holt and Marcie Traballoni for their technical expertise running the various versions of the TSI assay.
M. P., G. G. and A. W. M. are ARUP Institute employees.