Model-Based Analysis of In Vivo Release Data of Levonorgestrel Implants: Projecting Long-Term Systemic Exposure.

Levonorgestrel (LNG) is a progestin used in many contraceptive formulations, including subcutaneous implants. There is an unmet need for developing long-acting formulations for LNG. To develop long-acting formulations, release functions need to be investigated for LNG implant. Therefore, a release model was developed and integrated into an LNG physiologically-based pharmacokinetic (PBPK) model. Utilizing a previously developed LNG PBPK model, subcutaneous administration of 150 mg LNG was implemented into the modeling framework. To mimic LNG release, ten functions incorporating formulation-specific mechanisms were explored. Release kinetic parameters and bioavailability were optimized using Jadelle® clinical trial data (n = 321) and verified using two additional clinical trials (n = 216). The First-order release and Biexponential release models showed the best fit with observed data, the adjusted R-squared (R2) value is 0.9170. The maximum released amount is approximately 50% of the loaded dose and the release rate is 0.0009 per day. The Biexponential model also showed good agreement with the data (adjusted R2 = 0.9113). Both models could recapitulate observed plasma concentrations after integration into the PBPK simulations. First-order and Biexponential release functionality may be useful in modeling subcutaneous LNG implants. The developed model captures central tendency of the observed data as well as variability of release kinetics. Future work focuses on incorporating various clinical scenarios into model simulations, including drug-drug interactions and a range of BMIs.

A new method for successful indirect bonding in relation to bond strength.

The aim of the work was to develop a new transfer method for indirect bonding of brackets to improve the bond strength by applying a uniform contact pressure over the entire dental arch. This has a great potential to reduce the bracket loss rate during clinical treatment. A suitable shape memory polymer (SMP) was selected and prepared in the chemistry laboratory. This SMP applies a force to the brackets during bonding and thus increases the bond strength by applying uniform contact pressure. Various transfer trays were equipped with SMP platelets and the transfer of brackets from the plaster model to the real human tooth model was performed in vitro. The transfer accuracy and bond strength of the bonded brackets were investigated by 3D-overlay and shear tests, respectively. The transfer accuracy was technique sensitive and showed higher accuracy for the trays with SMPs and self-curing silicones than for the vacuum formed trays with SMPs. The bond strength of the indirectly bonded brackets with SMPs was on average 1-2 MPa higher than the bond strength of the brackets indirectly bonded with a conventional two-layer vacuum formed tray without SMPs. Thus, transfer trays with SMPs can provide a significant improvement in bond strength during indirect bonding after appropriate adjustment.

Development of a Translational Exposure-Bracketing Approach to Streamline the Development of Hormonal Contraceptive Drug Products.

Worldwide, 922 million women of reproductive age (or their partners) use some sort of contraception to prevent pregnancy. Oral combined hormonal contraceptives (CHCs) typically utilize a combination of a progestin and an estrogen. CHCs are potentially at risk to metabolic drug-drug interaction (DDI) via CYP3A4, the main enzyme involved in the oxidative metabolism of ethinyl estradiol and most progestins (e.g., levonorgestrel (LNG) and drospirenone (DRSP)). Recently, the US Food and Drug Administration (FDA) issued a guidance addressing metabolic DDIs in the realm of CHC, establishing an overall class-based recommendation with respect to avoidance of CYP3A4 induction interactions. Given that different progestins have varying magnitudes of fraction metabolized by CYP3A4 (fmCYP3A4 ), it would be of clinical benefit to determine if all progestins are at the same risk to CYP3A4-mediated metabolic DDIs. LNG and DRSP are commonly used progestins that are at the margins of the rifampicin induction effect observed in vivo because they have the relatively lowest and highest fmCYP3A4 among commonly used CHC formulations containing norgestimate, desogestrel, norgestrel, and norethindrone. Therefore, we applied a multi-pronged strategy (i.e., (i) development of the physiologically-based pharmacokinetic models; (ii) comparison of the effect of CYP3A inducers and inhibitors on DRSP vs. LNG; and (iii) providing the clinical-practice context based on real-world data, to explore the difference in DDI risk for oral CHCs.

Physiologically-based pharmacokinetic modeling to predict CYP3A4-mediated drug-drug interactions of finerenone.

Finerenone is a nonsteroidal, selective mineralocorticoid receptor antagonist that recently demonstrated its efficacy to delay chronic kidney disease (CKD) progression and reduce cardiovascular events in patients with CKD and type 2 diabetes. Here, we report the development of a physiologically-based pharmacokinetic (PBPK) model for finerenone and its application as a victim drug of cytochrome P450 3A4 (CYP3A4)-mediated drug-drug interactions (DDIs) using the open-source PBPK platform PK-Sim, which has recently been qualified for this application purpose. First, the PBPK model for finerenone was developed using physicochemical, in vitro, and clinical (including mass balance) data. Subsequently, the finerenone model was validated regarding the contribution of CYP3A4 metabolism to total clearance by comparing to observed data from dedicated clinical interaction studies with erythromycin (simulated geometric mean ratios of the area under the plasma concentration-time curve [AUCR] of 3.46 and geometric mean peak plasma concentration ratios [Cmax Rs] of 2.00 vs. observed of 3.48 and 1.88, respectively) and verapamil (simulated AUCR of 2.91 and Cmax R of 1.86 vs. observed of 2.70 and 2.22, respectively). Finally, the finerenone model was applied to predict clinically untested DDI studies with various CYP3A4 modulators. An AUCR of 6.31 and a Cmax R of 2.37 was predicted with itraconazole, of 5.28 and 2.25 with clarithromycin, 1.59 and 1.40 with cimetidine, 1.57 and 1.38 with fluvoxamine, 0.19 and 0.32 with efavirenz, and 0.07 and 0.14 with rifampicin. This PBPK analysis provides a quantitative basis to guide the label and clinical use of finerenone with concomitant CYP3A4 modulators.

Determining the Exposure Threshold for Levonorgestrel Efficacy Using an Integrated Model Based Meta-Analysis Approach.

Combined oral contraceptive pills are the most commonly used hormonal contraceptives for the prevention of unintended pregnancies in United States. They consist of a progestin (e.g., levonorgestrel (LNG)) and an estrogen component, typically ethinyl estradiol (EE). In addition to adherence issues, drug-drug interactions (DDIs) and obesity (women with body mass index (BMI) ≥ 30 kg/m2 ) are prime suspects for decreased LNG efficacy. Therefore, we developed an integrated physiologically-based pharmacokinetic modeling and model-based meta-analysis approach to determine LNG's efficacy threshold concentrations and to evaluate the impact of DDIs and obesity on the efficacy of LNG-containing hormonal contraceptives (HCs). Based on this approach, co-administration of strong CYP3A4 inducers and LNG-containing HCs (LNG150: LNG 150 µg + EE 30 µg and LNG100: LNG 100 µg + EE 20 µg) resulted in a predicted clinically relevant decrease of LNG plasma exposure (women with BMI < 25 kg/m2 : 50-65%; obese women: 70-75%). Following administration of LNG150 or LNG100 in the presence of a CYP3A4 inducer, there was an increase in mean Pearl Index of 1.2-1.30 and 1.80-2.10, respectively, in women with BMI < 25 kg/m2 (incidence rate ratios (IRRs): 1.7-2.2), whereas it ranged from 1.6-1.80 and 2.40-2.85 in obese women (IRR: 2.2-3.0), respectively. Our results suggest that the use of backup or alternate methods of contraception is not necessarily required for oral LNG + EE formulations except within circumstances of both obesity and strong CYP3A4 inducer concomitance following administration of LNG100.

Predictive Performance of Physiology-Based Pharmacokinetic Dose Estimates for Pediatric Trials: Evaluation With 10 Bayer Small-Molecule Compounds in Children.

Development and guidance of dosing schemes in children have been supported by physiology-based pharmacokinetic (PBPK) modeling for many years. PBPK models are built on a generic basis, where compound- and system-specific parameters are separated and can be exchanged, allowing the translation of these models from adults to children by accounting for physiological differences. Owing to these features, PBPK modeling is a valuable approach to support clinical decision making for dosing in children. In this analysis, we evaluate pediatric PBPK models for 10 small-molecule compounds that were applied to support clinical decision processes at Bayer for their predictive power in different age groups. Ratios of PBPK-predicted to observed PK parameters for the evaluated drugs in different pediatric age groups were estimated. Predictive performance was analyzed on the basis of a 2-fold error range and the bioequivalence range (ie, 0.8 ≤ predicted/observed ≤ 1.25). For all 10 compounds, all predicted-to-observed PK ratios were within a 2-fold error range (n = 27), with two-thirds of the ratios within the bioequivalence range (n = 18). The findings demonstrate that the pharmacokinetics of these compounds was successfully and adequately predicted in different pediatric age groups. This illustrates the applicability of PBPK for guiding dosing schemes in the pediatric population.

A generic framework for the physiologically-based pharmacokinetic platform qualification of PK-Sim and its application to predicting cytochrome P450 3A4-mediated drug-drug interactions.

The success of applications of physiologically-based pharmacokinetic (PBPK) modeling in drug development and drug labeling has triggered regulatory agencies to demand rigorous demonstration of the predictive capability of the specific PBPK platform for a particular intended application purpose. The effort needed to comply with such qualification requirements exceeds the costs for any individual PBPK application. Because changes or updates of a PBPK platform would require (re-)qualification, a reliable and efficient generic qualification framework is needed. We describe the development and implementation of an agile and sustainable technical framework for automatic PBPK platform (re-)qualification of PK-Sim® embedded in the open source and open science GitHub landscape of Open Systems Pharmacology. The qualification approach enables the efficient assessment of all aspects relevant to the qualification of a particular purpose and provides transparency and traceability for all stakeholders. As a showcase example for the power and versatility of the qualification framework, we present the qualification of PK-Sim® for the intended purpose of predicting cytochrome P450 3A4 (CYP3A4)-mediated drug-drug interactions (DDIs). Several perpetrator PBPK models featuring various degrees of CYP3A4 modulation and different types of mechanisms (competitive inhibition, mechanism-based inactivation, and induction) were coupled with a set of PBPK models of sensitive CYP3A4 victim drugs. Simulations were compared to a comprehensive data set of 135 observations from published clinical DDI studies. The platform's overall predictive performance showed reasonable accuracy and precision (geometric mean fold error of 1.4 for both area under the plasma concentration-time curve ratios and peak plasma concentration ratios with/without perpetrator) and suggests that PK-Sim® can be applied to quantitatively assess CYP3A4-mediated DDI in clinically untested scenarios.

Quantitative Assessment of Levonorgestrel Binding Partner Interplay and Drug-Drug Interactions Using Physiologically Based Pharmacokinetic Modeling.

Levonorgestrel (LNG) is the active moiety in many hormonal contraceptive formulations. It is typically coformulated with ethinyl estradiol (EE) to decrease intermenstrual bleeding. Due to its widespread use and CYP3A4-mediated metabolism, there is concern regarding drug-drug interactions (DDIs), particularly a suboptimal LNG exposure when co-administered with CYP3A4 inducers, potentially leading to unintended pregnancies. The goal of this analysis was to determine the impact of DDIs on the systemic exposure of LNG. To this end, we developed and verified a physiologically-based pharmacokinetic (PBPK) model for LNG in PK-Sim (version 8.0) accounting for the impact of EE and body mass index (BMI) on LNG's binding to sex-hormone binding globulin. Model parameters were optimized following intravenous and oral administration of 0.09 mg LNG. The combined LNG-EE PBPK model was verified regarding CYP3A4-mediated interaction by comparing to published clinical DDI study data with carbamazepine, rifampicin, and efavirenz (CYP3A4 inducers). Once verified, the model was applied to predict systemic LNG exposure in normal BMI and obese women (BMI ≥ 30 kg/m2 ) with and without co-administration of itraconazole (competitive CYP3A4 inhibitor) and clarithromycin (mechanism-based CYP3A4 inhibitor). Total and free LNG exposures, when co-administered with EE, decreased 2-fold in the presence of rifampin, whereas they increased 1.5-fold in the presence of itraconazole. Although changes in total and unbound exposure were decreased in obese women compared with normal BMI women, the relative impact of DDIs on LNG exposure was similar between both groups.

Application of Physiologically-Based and Population Pharmacokinetic Modeling for Dose Finding and Confirmation During the Pediatric Development of Moxifloxacin.

Moxifloxacin is a widely used fluoroquinolone for the treatment of complicated intra-abdominal infections. We applied physiologically-based pharmacokinetic (PBPK) and population pharmacokinetic (popPK) modeling to support dose selection in pediatric patients. We scaled an existing adult PBPK model to children based on prior physiological knowledge. The resulting model proposed an age-dependent dosing regimen that was tested in a phase I study. Refined doses were then tested in a phase III study. A popPK analysis of all clinical pediatric data confirmed the PBPK predictions, including the proposed dosing schedule in children, and supported pharmacokinetics-related safety/efficacy questions. The pediatric PBPK model adequately predicted the doses necessary to achieve antimicrobial efficacy while maintaining safety in the phase I and III pediatric studies. Altogether, this study retroactively demonstrated the robustness and utility of modeling to support dose finding and confirmation in pediatric drug development for moxifloxacin.

Physiologically Based Pharmacokinetic Models for Prediction of Complex CYP2C8 and OATP1B1 (SLCO1B1) Drug-Drug-Gene Interactions: A Modeling Network of Gemfibrozil, Repaglinide, Pioglitazone, Rifampicin, Clarithromycin and Itraconazole.

BACKGROUND: Drug-drug interactions (DDIs) and drug-gene interactions (DGIs) pose a serious health risk that can be avoided by dose adaptation. These interactions are investigated in strictly controlled setups, quantifying the effect of one perpetrator drug or polymorphism at a time, but in real life patients frequently take more than two medications and are very heterogenous regarding their genetic background. OBJECTIVES: The first objective of this study was to provide whole-body physiologically based pharmacokinetic (PBPK) models of important cytochrome P450 (CYP) 2C8 perpetrator and victim drugs, built and evaluated for DDI and DGI studies. The second objective was to apply these models to describe complex interactions with more than two interacting partners. METHODS: PBPK models of the CYP2C8 and organic-anion-transporting polypeptide (OATP) 1B1 perpetrator drug gemfibrozil (parent-metabolite model) and the CYP2C8 victim drugs repaglinide (also an OATP1B1 substrate) and pioglitazone were developed using a total of 103 clinical studies. For evaluation, these models were applied to predict 34 different DDI studies, establishing a CYP2C8 and OATP1B1 PBPK DDI modeling network. RESULTS: The newly developed models show a good performance, accurately describing plasma concentration-time profiles, area under the plasma concentration-time curve (AUC) and maximum plasma concentration (Cmax) values, DDI studies as well as DGI studies. All 34 of the modeled DDI AUC ratios (AUC during DDI/AUC control) and DDI Cmax ratios (Cmax during DDI/Cmax control) are within twofold of the observed values. CONCLUSIONS: Whole-body PBPK models of gemfibrozil, repaglinide, and pioglitazone have been built and qualified for DDI and DGI prediction. PBPK modeling is applicable to investigate complex interactions between multiple drugs and genetic polymorphisms.

Physiologically-Based Pharmacokinetic Models for CYP1A2 Drug-Drug Interaction Prediction: A Modeling Network of Fluvoxamine, Theophylline, Caffeine, Rifampicin, and Midazolam.

This study provides whole-body physiologically-based pharmacokinetic models of the strong index cytochrome P450 (CYP)1A2 inhibitor and moderate CYP3A4 inhibitor fluvoxamine and of the sensitive CYP1A2 substrate theophylline. Both models were built and thoroughly evaluated for their application in drug-drug interaction (DDI) prediction in a network of perpetrator and victim drugs, combining them with previously developed models of caffeine (sensitive index CYP1A2 substrate), rifampicin (moderate CYP1A2 inducer), and midazolam (sensitive index CYP3A4 substrate). Simulation of all reported clinical DDI studies for combinations of these five drugs shows that the presented models reliably predict the observed drug concentrations, resulting in seven of eight of the predicted DDI area under the plasma curve (AUC) ratios (AUC during DDI/AUC control) and seven of seven of the predicted DDI peak plasma concentration (Cmax ) ratios (Cmax during DDI/Cmax control) within twofold of the observed values. Therefore, the models are considered qualified for DDI prediction. All models are comprehensively documented and publicly available, as tools to support the drug development and clinical research community.

PBPK Models for CYP3A4 and P-gp DDI Prediction: A Modeling Network of Rifampicin, Itraconazole, Clarithromycin, Midazolam, Alfentanil, and Digoxin.

According to current US Food and Drug Administration (FDA) and European Medicines Agency (EMA) guidance documents, physiologically based pharmacokinetic (PBPK) modeling is a powerful tool to explore and quantitatively predict drug-drug interactions (DDIs) and may offer an alternative to dedicated clinical trials. This study provides whole-body PBPK models of rifampicin, itraconazole, clarithromycin, midazolam, alfentanil, and digoxin within the Open Systems Pharmacology (OSP) Suite. All models were built independently, coupled using reported interaction parameters, and mutually evaluated to verify their predictive performance by simulating published clinical DDI studies. In total, 112 studies were used for model development and 57 studies for DDI prediction. 93% of the predicted area under the plasma concentration-time curve (AUC) ratios and 94% of the peak plasma concentration (Cmax ) ratios are within twofold of the observed values. This study lays a cornerstone for the qualification of the OSP platform with regard to reliable PBPK predictions of enzyme-mediated and transporter-mediated DDIs during model-informed drug development. All presented models are provided open-source and transparently documented.

Clarithromycin, Midazolam, and Digoxin: Application of PBPK Modeling to Gain New Insights into Drug-Drug Interactions and Co-medication Regimens.

Clarithromycin is a substrate and mechanism-based inhibitor of cytochrome P450 (CYP) 3A4 as well as a substrate and competitive inhibitor of P-glycoprotein (P-gp) and organic anion-transporting polypeptides (OATP) 1B1 and 1B3. Administered concomitantly, clarithromycin causes drug-drug interactions (DDI) with the victim drugs midazolam (CYP3A4 substrate) and digoxin (P-gp substrate). The objective of the presented study was to build a physiologically based pharmacokinetic (PBPK) DDI model for clarithromycin, midazolam, and digoxin and to exemplify dosing adjustments under clarithromycin co-treatment. The PBPK model development included an extensive literature search for representative PK studies and for compound characteristics of clarithromycin, midazolam, and digoxin. Published concentration-time profiles were used for model development (training dataset), and published and unpublished individual profiles were used for model evaluation (evaluation dataset). The developed single-compound PBPK models were linked for DDI predictions. The full clarithromycin DDI model successfully predicted the metabolic (midazolam) and transporter (digoxin) DDI, the acceptance criterion (0.5 ≤ AUCratio,predicted/AUCratio,observed ≤ 2) was met by all predictions. During co-treatment with 250 or 500 mg clarithromycin (bid), the midazolam and digoxin doses should be reduced by 74 to 88% and by 21 to 22%, respectively, to ensure constant midazolam and digoxin exposures (AUC). With these models, we provide highly mechanistic tools to help researchers understand and characterize the DDI potential of new molecular entities and inform the design of DDI studies with potential CYP3A4 and P-gp substrates.

Development of a physiologically based computational kidney model to describe the renal excretion of hydrophilic agents in rats.

A physiologically based kidney model was developed to analyze the renal excretion and kidney exposure of hydrophilic agents, in particular contrast media, in rats. In order to study the influence of osmolality and viscosity changes, the model mechanistically represents urine concentration by water reabsorption in different segments of kidney tubules and viscosity dependent tubular fluid flow. The model was established using experimental data on the physiological steady state without administration of any contrast media or drugs. These data included the sodium and urea concentration gradient along the cortico-medullary axis, water reabsorption, urine flow, and sodium as well as urea urine concentrations for a normal hydration state. The model was evaluated by predicting the effects of mannitol and contrast media administration and comparing to experimental data on cortico-medullary concentration gradients, urine flow, urine viscosity, hydrostatic tubular pressures and single nephron glomerular filtration rate. Finally the model was used to analyze and compare typical examples of ionic and non-ionic monomeric as well as non-ionic dimeric contrast media with respect to their osmolality and viscosity. With the computational kidney model, urine flow depended mainly on osmolality, while osmolality and viscosity were important determinants for tubular hydrostatic pressure and kidney exposure. The low diuretic effect of dimeric contrast media in combination with their high intrinsic viscosity resulted in a high viscosity within the tubular fluid. In comparison to monomeric contrast media, this led to a higher increase in tubular pressure, to a reduction in glomerular filtration rate and tubular flow and to an increase in kidney exposure. The presented kidney model can be implemented into whole body physiologically based pharmacokinetic models and extended in order to simulate the renal excretion of lipophilic drugs which may also undergo active secretion and reabsorption.

A computational systems biology software platform for multiscale modeling and simulation: integrating whole-body physiology, disease biology, and molecular reaction networks.

Today, in silico studies and trial simulations already complement experimental approaches in pharmaceutical R&D and have become indispensable tools for decision making and communication with regulatory agencies. While biology is multiscale by nature, project work, and software tools usually focus on isolated aspects of drug action, such as pharmacokinetics at the organism scale or pharmacodynamic interaction on the molecular level. We present a modeling and simulation software platform consisting of PK-Sim(®) and MoBi(®) capable of building and simulating models that integrate across biological scales. A prototypical multiscale model for the progression of a pancreatic tumor and its response to pharmacotherapy is constructed and virtual patients are treated with a prodrug activated by hepatic metabolization. Tumor growth is driven by signal transduction leading to cell cycle transition and proliferation. Free tumor concentrations of the active metabolite inhibit Raf kinase in the signaling cascade and thereby cell cycle progression. In a virtual clinical study, the individual therapeutic outcome of the chemotherapeutic intervention is simulated for a large population with heterogeneous genomic background. Thereby, the platform allows efficient model building and integration of biological knowledge and prior data from all biological scales. Experimental in vitro model systems can be linked with observations in animal experiments and clinical trials. The interplay between patients, diseases, and drugs and topics with high clinical relevance such as the role of pharmacogenomics, drug-drug, or drug-metabolite interactions can be addressed using this mechanistic, insight driven multiscale modeling approach.

Early developmental specification of the thyroid gland depends on han-expressing surrounding tissue and on FGF signals.

The thyroid is an endocrine gland in all vertebrates that develops from the ventral floor of the anterior pharyngeal endoderm. Unravelling the molecular mechanisms of thyroid development helps to understand congenital hypothyroidism caused by the absence or reduction of this gland in newborn humans. Severely reduced or absent thyroid-specific developmental genes concomitant with the complete loss of the functional gland in the zebrafish hands off (han, hand2) mutant reveals the han gene as playing a novel, crucial role in thyroid development. han-expressing tissues surround the thyroid primordium throughout development. Fate mapping reveals that, even before the onset of thyroid-specific developmental gene expression, thyroid precursor cells are in close contact with han-expressing cardiac lateral plate mesoderm. Grafting experiments show that han is required in surrounding tissue, and not in a cell-autonomous manner, for thyroid development. Loss of han expression in the branchial arches and arch-associated cells after morpholino knock-down of upstream regulator genes does not impair thyroid development, indicating that other han-expressing structures, most probably cardiac mesoderm, are responsible for the thyroid defects in han mutants. The zebrafish ace (fgf8) mutant has similar thyroid defects as han mutants, and chemical suppression of fibroblast growth factor (FGF) signalling confirms that this pathway is required for thyroid development. FGF-soaked beads can restore thyroid development in han mutants, showing that FGFs act downstream of or in parallel to han. These data suggest that loss of FGF-expressing tissue in han mutants is responsible for the thyroid defects.

Analysis of origin and growth of the thyroid gland in zebrafish.

The zebrafish thyroid gland shows a unique pattern of growth as a differentiated endocrine gland. Here, we analyze the onset of differentiation, the contribution of lineages, and the mode of growth of this gland. The expression of genes involved in hormone production and the establishment of epithelial polarity show that differentiation into a first thyroid follicle takes place early during embryonic development. Thyroid follicular tissue then grows along the pharyngeal midline, initially independently of thyroid stimulating hormone. Lineage analysis reveals that thyroid follicle cells are exclusively recruited from the pharyngeal endoderm. The ultimobranchial bodies that merge with the thyroid in mammals form separate glands in zebrafish as visualized by calcitonin precursor gene expression. Mosaic analysis suggests that the first thyroid follicle differentiating at 55 hours postfertilization corresponds later to the most anterior follicle and that new follicles are added caudally.

Pax2.1 is required for the development of thyroid follicles in zebrafish.

The thyroid gland is an organ primarily composed of endoderm-derived follicular cells. Although disturbed embryonic development of the thyroid gland leads to congenital hypothyroidism in humans and mammals, the underlying principles of thyroid organogenesis are largely unknown. In this study, we introduce zebrafish as a model to investigate the molecular and genetic mechanisms that control thyroid development. Marker gene expression suggests that the molecular pathways of early thyroid development are essentially conserved between fish and mammals. However during larval stages, we find both conserved and divergent features of development compared with mammals. A major difference is that in fish, we find evidence for hormone production not only in thyroid follicular cells, but also in an anterior non-follicular group of cells. We show that pax2.1 and pax8, members of the zebrafish pax2/5/8 paralogue group, are expressed in the thyroid primordium. Whereas in mice, only Pax8 has a function during thyroid development, analysis of the zebrafish pax2.1 mutant no isthmus (noi(-/-)) demonstrates that pax2.1 has a role comparable with mouse Pax8 in differentiation of the thyroid follicular cells. Early steps of thyroid development are normal in noi(-/-), but later expression of molecular markers is lost and the formation of follicles fails. Interestingly, the anterior non-follicular site of thyroid hormone production is not affected in noi(-/-). Thus, in zebrafish, some remaining thyroid hormone synthesis takes place independent of the pathway leading to thyroid follicle formation. We suggest that the noi(-/-) mutant serves as a new zebrafish model for hypothyroidism.