Screening for mutations was performed in a group of 314 patients from the Charité Pediatric Diabetes and Obesity outpatient clinic. Of these, 94 patients suffered from antibody-negative diabetes with onset before the age of 18 years and did not have mutations in genes causing monogenic diabetes (GCK, HNF4A, HNF1A, HNF1B, ABCC8, KCNJ11, INS). The remaining 220 patients were overweight or obese and exhibited impaired glucose homeostasis. Overweight was defined as >90. percentile and obesity >97. percentile according to German references (Kromeyer-Hauschild et al., 2001). Impaired glucose homeostasis was diagnosed by an increased insulin resistance in a fasting state, determined by elevated HOMA-IR (homeostatic model assessment in insulin resistance) or by an impaired oral glucose tolerance test.

The variant frequencies of p.ArgR23Cys (rs8192618) and p.Ser49Leu (rs140960896) were determined for 2,018 diabetes-free participants older than 60 years of age from the population-based SHIP and SHIP-TREND cohorts. Individuals were classified as diabetics if they were diagnosed accordingly (type I or II) by a physician (self-reported during an interview), were under anti-diabetic medication (ATC-code A10), or exhibited increased blood or serum glucose levels (HbA1c ≥ 6.5 and % ≥ 11.1 mmol/l, respectively). All other individuals were classified as diabetes-free.

Study of Health in Pomerania is a population-based project in West Pomerania, a region in the northeast of Germany, that consists of two independent prospectively collected cohorts (SHIP and SHIP-TREND) assessing the prevalence and incidence of common population-based diseases and their risk factors. The study design has been previously described in detail (Volzke et al., 2011).

For SHIP, baseline examinations were carried out from 1997 until 2001, and the sample finally comprised 4,308 participants. Baseline examinations for SHIP-TREND were carried out between 2008 and 2012 and comprised a total of 4420 participants.

The SHIP and SHIP-TREND samples were genotyped using the Illumina ExomeChip array (version Infinium HumanExome v1.0 DNA Analysis BeadChip). Array processing was performed in accordance with the manufacturer’s standard recommendations at the Helmholtz Zentrum München. Genotype calling was performed according to the ExomeChip quality control SOP version 5. Initial genotypes were determined using the GenomeStudio Genotyping Module v1.0 (GenCall algorithm) with the HumanExome-12v1_A manifest file and the standard Illumina cluster file (HumanExome-12v1.egt). Contaminated samples, samples with a call rate < 90%, extreme heterozygosity, extensive estimated IBD sharing with a large number of samples, or mismatch between reported and genotyped sex were excluded. Next, missing genotypes were recalled with the zCall software version 3.3 using default values. In both cohorts, a total of 8,177 individuals were successfully genotyped with an average call rate of 99.97%.

PCR amplification of the TAAR1 gene was performed by using the following primer pairs: forward tttcctcctaggtttctggga and reverse tccaccactgaacagctgac, located in the 5′ and 3′ region of the gene and giving rise to a 1457 bp long PCR fragment. Variant screening was performed by using the amplification primer and two additional primers (CTGGAGCTAAACTTCAAAGG and CTTGCCTGTTCTTTAGCGAT, located in the coding region of the gene) to cover the whole region in both directions using. For sequencing BigDye terminator (PerkinElmer Inc., Waltham, MA, United States) automatic sequencing (ABI3710xl, Applied Biosystems, Foster City, CA, United States) of the coding region of TAAR1 (NP_612200.1, NM_138327.2) was applied. TAAR1 wild-type (TAAR1-WT) was cloned in the eukaryotic expression vector pcDps as previously described (Kleinau et al., 2011).

The three single nucleotide variants identified by patient screening, p.Arg23Cys (R23C), p.Ser49Leu (S49L), and p.Ille171Leu (I171L), were introduced into TAAR1 using site-directed mutagenesis according to manufacturer’s instructions of Pfu turbo DNA polymerase (Stratagene, La Jolla, CA, United States).

To enhance cell surface expression for functional analysis, TAAR1 was N-terminally fused with the first 20 amino acids of the bovine rhodopsin (Rho-tag) as previously described (Dinter et al., 2015).

For surface ELISA (enzyme-linked immunosorbent assay), receptor constructs were N-terminally tagged with a hemagglutinin epitope (HA, 5′ YPYDVPDYA 3′) (previous to the Rho-tag).

For SNAP-tag technology, TAAR1-WT and variants were subcloned into the N-terminally SNAP-tagged pcDNA5 vector (New England Biolabs GmbH, Frankfurt am Main, Germany).

For cAMP assay and SNAP-tag experiments, HEK293 cells were cultured in Minimum Essential Media (MEM) and for cell surface ELISA, COS-7 cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (both Biochrom AG, Berlin, Germany), at 37°C and 5% CO₂. MEM was supplemented with 5% FBS and non-essential amino acids (Biochrom AG, Berlin, Germany). DMEM was supplemented with 5% FBS and 100 U/ml penicillin, 100 μg/ml streptomycin (Biochrom AG, Berlin, Germany), and 2 mM L-glutamine (Invitrogen, Carlsbad, CA, United States). We used two different cell lines because HEK293 cells are an established cell system for investigation of GPCR structure-function relationships. However, for methods that involve intensive washing, like the sandwich ELISA approach, a more robust cell line like COS-7 cells is appropriate.

Transient transfection was carried out using Metafectene (Biontex, Munich, Germany) for cAMP Assay and surface ELISA or FuGene6 (Promega, Fitchburg, WI, United States) for SNAP-tag experiments according to the manufacturers’ protocol.

For determination of cell surface expression with fluorescent-based SNAP-tag technology, cells expressing TAAR1-WT and mutant receptor constructs were fixed with 4% PFA and 200 mM HEPES followed by staining with the photostable fluorescent SNAP-Surface^® Alexa Fluor^® 488 (1:1,000 diluted in Advanced MEM (New England Biolabs GmbH, Frankfurt am Main, Germany), which covalently binds to the SNAP-tag (Cole, 2013).

For SNAP-tag technology nuclei were stained using DAPI dye (Life Technologies, Darmstadt, Germany). Microscopy was performed using an inverted light microscope Axiovert 10 (Carl Zeiss Microscopy GmbH, Oberkochen, Germany) at 63x magnification. Additionally, for verification and quantification of results, a surface ELISA was performed. Therefore, COS-7 cells were transiently transfected with the HA-tagged TAAR1 constructs and empty vector (mock, negative control).

For cell surface expression, cells were fixed with 4% paraformaldehyde and blocked with 10% FBS overnight 48 h after transfection, Next, cells were incubated with biotin-labeled anti-HA antibody (Roche, Basel, Switzerland) (1:200). Bound antibodies were detected with horseradish peroxidase-labeled Streptavidin (BioLegend, London, United Kingdom) (1:2,500). Color reaction was performed as described (Muhlhaus et al., 2014) and absorption was measured at 492 nm and 620 nm using an Anthos Reader 2001 (Anthos Labtech Instruments, Salzburg, Austria).

HEK293 cells transiently expressing empty vector (mock), TAAR1-WT, TAAR1-R23C, TAAR1-S49L, or TAAR1-I171L were left untreated or were stimulated with increasing concentrations of T1AM (from 1 nM to 10 μM) (Santa Cruz Biotechnology Inc., Dallas, TX, United States) or β-phenylethylamine (PEA) 10 μM in the presence of isobuthylmethylxanthine (IBMX). After cell lysis, cAMP levels were measured using AlphaScreen Technology (Perkin-Elmer Life Science, Boston, MA, United States) (Kleinau et al., 2011).

Statistical analyses were performed using the R statistical software version 2.15.3¹ and GraphPad Prism 6.0. One-way ANOVA was performed for the effect of TAAR1 stimulation by T1AM or PEA and for cell surface expression.

The details of structural TAAR1 homology modeling were described previously (Kleinau et al., 2011). In brief, the already determined structure of the β-adrenergic receptor 2 (pdb entry 2RH1, Cherezov et al., 2007) was used as a structural template for TAAR1 modeling in an inactive state, based on high sequence similarity between hTAAR1 and the β-adrenergic receptor 2. The loops were manually adjusted in length and amino acid composition. The ligand bound to the template was deleted. For all modeling procedures, the software package Sybyl 8.0 (Certara, NJ, United States) was used. Substituted side chains and loops were subjected to molecular dynamics simulation (4 ns) by fixing the backbone of the transmembrane helices, followed by conjugate gradient minimizations until converging at a termination gradient of 0.05 kcal/(mol^∗Å). The AMBER 7.0 force field was used. Structure images were produced using PyMOL Molecular Graphics System, version 1.5, Schrödinger, LLC.

Screening of 314 patients with impaired glucose tolerance revealed three heterozygous carriers of single nucleotide variants in the TAAR1 coding region: p.Arg23Cys (R23C, rs8192618), p.Ser49Leu (S49L, rs140960896), and p.Ile171Leu (I171L, rs200795344). This corresponds to a minor allele frequency (MAF) of 0.16% for each variant in this patient sample.

Two different methods were used to investigate cell surface localization of TAAR1. The SNAP technology revealed TAAR1 wild-type (TAAR1-WT) and variants in intact cells (Figure 1), whereas the cell surface ELISA allowed for the quantification of cell surface expression (Figure 2A). Using both methods, cell surface localization of all three mutated TAAR1 proteins was statistically not different to wild-type.

Wild-type TAAR1 and the three TAAR1 mutant proteins were expressed in HEK293 cells and T1AM-induced cAMP accumulation was measured. Compared with wild-type TAAR1 (EC₅₀: 2.8 μM), the R23C variant demonstrated a complete loss of function in cAMP accumulation (significant, p ≤ 0.001), which did not allow for a proper EC₅₀ calculation. Maximal stimulation for S49L and I171L was reduced to 60% (significant, p ≤ 0.05) and 80% (not significant) compared with the wild-type, respectively (Figure 3A).

To mimic the heterozygous genotype of the patients, co-transfection studies of wild-type (WT) and mutant receptors were performed. For determination of whether cell surface localization of the WT receptor is influenced by the mutated receptors, cells were co-transfected with Flag-tagged TAAR1-WT and HA-tagged TAAR1 variants. While cell surface localization was not impaired for all three variants (Figure 2B), R23C and S49L co-expression with TAAR1-WT significantly reduced maximal Gs signaling to 58% (significant, p ≤ 0.05), and 55% (significant, p ≤ 0.05) of WT + empty vector signaling, respectively (Figure 3B), indicating a dominant-negative effect. In contrast, signaling of the TAAR1 I177L variant + TAAR1-WT was not affected. Furthermore, the S49L variant resulted in a slight shift toward higher EC₅₀ values (0.7 μM for TAAR1 + empty vector and 1.9 μM for wild-type + S49L) (Figure 3B).

Beta-phenylethylamine (PEA) is known to be one of the most potent ligands for TAAR1 (Wainscott et al., 2007). For determination of Gs signaling properties after stimulation with PEA, HEK293 cells transiently transfected with TAAR1 wild-type (TAAR1-WT) and TAAR1 variants were stimulated with 10^-5M PEA and cAMP accumulation was measured. The R23C mutant showed a complete loss of function indicated by non-responsiveness to PEA (significant, p ≤ 0.001) (Figure 4). The Gs/adenylyl cyclase-mediated signaling of the S49L variant demonstrated a significant signal reduction of approximately 70%, (p ≤ 0.001) while the I171L substitution exhibited only a slight, but not significant, reduction of cAMP accumulation.

In summary, R23C represented a complete loss-of-function variant, S49L a partial loss-of-function variant, and despite a slightly reduced maximal signaling capacity, I171L was classified as WT-like and excluded from further analysis.

Patient 1 (variant p.Arg23Cys carrier) was born small for gestational age to non-related parents of Middle Eastern background. He developed insulin-dependent diabetes at 3 years of age with low C-peptide, and negative anti-GAD and anti-insulin antibodies. Further evaluation did not reveal variants in known monogenic diabetes genes. At 7 years of age, his mental performance was tested and revealed an IQ of 71. Additionally, his growth rate consistently decreased and at 6 years of age, he was tested for growth hormone deficiency with normal MRI findings of the brain and pituitary gland. His father (non-carrier of a variant) developed type 2 diabetes at 40 years of age and was not obese. Treatment was comprised of oral antidiabetic drugs. The mother is a heterozygous variant carrier and is normoglycemic. Thyroid hormone levels were in the reference range.

Patient 2 (variant p.Ile171Leu carrier) was diagnosed at 11 4/12 years of age with obesity (BMI, 30.0 kg/m²; +2.59 BMI-SDS) and a disturbed glucose homeostasis (HOMA-IR, 2.8) and transient pathological oral glucose tolerance test. There was no family background of type 2 diabetes or cardiovascular problems. Lifestyle intervention resulted in weight loss (BMI, 25.5 kg/m²) and HOMA-IR values normalized but were not completely stable (0.7–1.7). Thyroid function was normal.

Patient 3 (variant p.Ser49Leu carrier) was overweight (BMI 25.0 kg/m²) and had a disturbed glucose homeostasis (IGT; HOMA-IR, 2.9). Lifestyle intervention resulted in weight loss and normalization of HOMA-IR; however, the patient regained weight (BMI, 29.0 kg/m²) and HOMA-IR increased to 6. There was no reported family history of type 2 diabetes or cardiovascular complications. Thyroid function was normal. The patient discontinued from the treatment group because of psychiatric problems.

Samples from the parents of patients 2 and 3 were unavailable.

The results so far suggest that two of the three single nucleotide variants, namely p.Arg23Cys (rs8192618) and p.Ser49Leu (rs140960896), might predispose to impaired glucose homeostasis. The design of a proper control cohort to detect putative enrichment of these variants in patients had to anticipate a variable and mild manifestation of the phenotype with increasing age. Therefore, participants of the SHIP study that were ≥60 years of age and diabetes-free were assigned to this control group, thereby increasing the likelihood of including unaffected individuals. Within the SHIP and SHIP-TREND cohorts, 2,018 individuals that were ≥60 years of age were diagnosed as diabetes-free (see description of the control population) and were subsequently analyzed for the p.Arg23Cys (rs8192618, genotyping data for 2,009 individuals available) and p.Ser49Leu (rs140960896, genotyping data for 2,018 individuals available) allele frequencies. In this control group, only nine individuals were found to be heterozygous carriers of p.Ser49Leu, while no carriers were observed for p.Arg23Cys (Table 1). In the remaining SHIP sample, patients with <60 years of age that included individuals with the potential of developing diabetes at an older age, the number of heterozygous p.Ser49Leu carriers amounted to 18 of 5,149 individuals with available genotyping data. Four heterozygous p.Arg23Cys carriers were found among the 5,163 individuals with the available genotyping data. The corresponding single nucleotide variant minor allele frequencies (MAFs) for the older control cohort, the younger remaining cohort, and the total SHIP study were 0.00%, 0.04%, and 0.03% for p.Arg23Cys (rs8192618), respectively, and 0.22%, 0.17%, and 0.19% for p.Ser49Leu (rs140960896), respectively. The MAFs reported for the aggregated general population (121,408 samples from the dbSNP database²) amount to 0.154 and 0.038% for p.Arg23Cys and p.Ser49Leu, respectively.