Application of machine learning methods to bridge the gap between non-interventional studies and randomized controlled trials in ophthalmic patients with neovascular age-related macular degeneration.

PURPOSE: The effectiveness of intravitreal anti-vascular endothelial growth factor agents is usually lower in real world settings compared with randomized clinical trials (RCTs), often limiting the use of real-world evidence (RWE) in regulatory and healthcare decisions. The current analysis aimed to develop and validate an algorithm to explain the difference in outcomes between RWE studies and RCTs in patients with neovascular age-related macular degeneration. METHODS: The algorithm was developed using ranibizumab real world data (RWD) from the US and validated on Australian and UK RWD. A decision model was developed using machine learning principles, in which the model learns how to partition the most influential factors (out of 59 variables) so that they maximally relate to the change in visual acuity (VA) over 12 months. RESULTS: The algorithm identified baseline VA <73 Early Treatment Diabetic Retinopathy Study letters, presence of baseline subretinal fluid, and administration of three loading doses by Day 90 from drug initiation as the characteristics with the greatest impact on VA at month 12. When applying the different criteria, RWE outcomes became similar to those obtained in known RCTs. CONCLUSION: Machine learning techniques can be used to classify real world cohorts and identify subsets of patients who benefit to the same extent as that reported in RCTs. This methodology may support the translation of clinical trial findings to treatment performance in the clinical practice setting.

Pharmacological estrogen administration causes a FSH-independent osteo-anabolic effect requiring ER alpha in osteoblasts.

Postmenopausal osteoporosis is characterized by declining estrogen levels, and estrogen replacement therapy has been proven beneficial for preventing bone loss in affected women. While the physiological functions of estrogen in bone, primarily the inhibition of bone resorption, have been studied extensively, the effects of pharmacological estrogen administration are still poorly characterized. Since elevated levels of follicle-stimulating hormone (FSH) have been suggested to be involved in postmenopausal bone loss, we investigated whether the skeletal response to pharmacological estrogen administration is mediated in a FSH-dependent manner. Therefore, we treated wildtype and FSHß-deficicent (Fshb(-/-)) mice with estrogen for 4 weeks and subsequently analyzed their skeletal phenotype. Here we observed that estrogen treatment resulted in a significant increase of trabecular and cortical bone mass in both, wildtype and Fshb(-/-) mice. Unexpectedly, this FSH-independent pharmacological effect of estrogen was not caused by influencing bone resorption, but primarily by increasing bone formation. To understand the cellular and molecular nature of this osteo-anabolic effect we next administered estrogen to mouse models carrying cell specific mutant alleles of the estrogen receptor alpha (ERα). Here we found that the response to pharmacological estrogen administration was not affected by ERα inactivation in osteoclasts, while it was blunted in mice lacking the ERα in osteoblasts or in mice carrying a mutant ERα incapable of DNA binding. Taken together, our findings reveal a previously unknown osteo-anabolic effect of pharmacological estrogen administration, which is independent of FSH and requires DNA-binding of ERα in osteoblasts.

Dynamics of the transcriptome in the primate ovulatory follicle.

Experiments were designed to evaluate changes in the transcriptome (mRNA levels) in the ovulatory, luteinizing follicle of rhesus monkeys, using a controlled ovulation model that permits analysis of the naturally selected, dominant follicle at specific intervals (0, 12, 24 and 36 h) after exposure to an ovulatory (exogenous hCG) stimulus during the menstrual cycle. Total RNA was prepared from individual follicles (n= 4-8/timepoint), with an aliquot used for microarray analysis (Affymetrix Rhesus Macaque Genome Array) and the remainder applied to quantitative real-time PCR (q-PCR) assays. The microarray data from individual samples distinctly clustered according to timepoints, and ovulated follicles displayed markedly different expression patterns from unruptured follicles at 36 h. Between timepoint comparisons revealed profound changes in mRNA expression profiles. The dynamic pattern of mRNA expression for steroidogenic enzymes (CYP17A, CYP19A, HSD3B2, HSD11B1 and HSD11B2), steroidogenic acute regulatory protein (StAR) and gonadotrophin receptors [LH/choriogonadotrophin receptor (LHCGR), FSH receptor (FSHR)] as determined by microarray analysis correlated precisely with those from blinded q-PCR assays. Patterns of mRNA expression for epidermal-growth-factor-like factors (amphiregulin, epiregulin) and processes [hyaluronan synthase 2 (HAS2), tumor necrosis factor alpha-induced protein 6 (TNFAIP6)] implicated in cumulus-oocyte maturation/expansion were also comparable between assays. Thus, several mRNAs displayed the expected expression pattern for purported theca (e.g. CYP17A), granulosa (CYP19A, FSHR), cumulus (HAS2, TNFAIP6) cell and surface epithelium (HSD11B)-related genes in the rodent/primate pre-ovulatory follicle. This database will be of great value in analyzing molecular and cellular pathways associated with periovulatory events in the primate follicle (e.g. follicle rupture, luteinization, inflammatory response and angiogenesis), and for identifying novel gene products controlling mammalian fertility.

Neuronal estrogen receptor-alpha mediates neuroprotection by 17beta-estradiol.

17beta-Estradiol (E(2)) was shown to exert neuroprotective effects both in in vitro and in vivo models of stroke. Although these effects of E(2) are known to require estrogen receptor-alpha (ER alpha), the cellular target of estrogen-mediated neuroprotection remains unknown. Using cell type-specific ER mutant mice in an in vivo model of stroke, we specifically investigated the role of ER alpha in neuronal cells versus its role in the microglia in the mediation of neuroprotection by estrogens. We generated and analyzed two different tissue-specific knockout mouse lines lacking ER alpha either in cells of myeloid lineage, including microglia, or in the neurons of the forebrain. Both E(2)-treated and E(2)-untreated mutant and control mice were subjected to a permanent middle cerebral artery occlusion for 48 h, and the infarct volume was quantified. Although the infarct volume of E(2)-treated female myeloid-specific ER alpha knockout mice was similar to that of E(2)-treated control mice, both male and female neuron-specific ER alpha mutant mice had larger infarcts than did control mice after E(2) treatment. We conclude that neuronal ER alpha in female and male mice mediates neuroprotective estrogen effects in an in vivo mouse model of stroke, whereas microglial ER alpha is dispensable.

DNA binding by estrogen receptor-alpha is essential for the transcriptional response to estrogen in the liver and the uterus.

The majority of the biological effects of estrogens in the reproductive tract are mediated by estrogen receptor (ER)alpha, which regulates transcription by several mechanisms. Because the tissue-specific effects of some ERalpha ligands may be caused by tissue-specific transcriptional mechanisms of ERalpha, we aimed to identify the contribution of DNA recognition to these mechanisms in two clinically important target organs, namely uterus and liver. We used a genetic mouse model that dissects DNA binding-dependent vs. independent transcriptional regulation elicited by ERalpha. The EAAE mutant harbors amino acid exchanges at four positions of the DNA-binding domain (DBD) of ERalpha. This construct was knocked in the ERalpha gene locus to produce ERalpha((EAAE/EAAE)) mice devoid of a functional ERalpha DBD. The phenotype of the ERalpha((EAAE/EAAE)) mice resembles the general loss-of-function phenotype of alphaER knockout mutant mice with hypoplastic uteri, hemorrhagic ovaries, and impaired mammary gland development. In agreement with this phenotype, the expression pattern of the ERalpha((EAAE/EAAE)) mutant mice in liver obtained by genome-wide gene expression profiling supports the observation of a near-complete loss of estrogen-dependent gene regulation in comparison with the wild type. Further gene expression analyses to validate the results of the microarray data were performed by quantitative RT-PCR. The analyses indicate that both gene activation and repression by estrogen-bound ERalpha rely on an intact DBD in vivo.

Inactivation of the gene for the nuclear receptor tailless in the brain preserving its function in the eye.

During embryogenesis, tailless, an orphan member of the nuclear receptor family, is expressed in the germinal zones of the brain and the developing retina, and is involved in regulating the cell cycle of progenitor cells. Consequently, a deletion of the tailless gene leads to decreased cell number with associated anatomical defects in the limbic system, the cortex and the eye. These structural abnormalities are associated with blindness, increased aggressiveness, poor performance in learning paradigms and reduced anxiousness. In order to assess the contribution of blindness to the behavioural changes, we established tailless mutant mice with intact visual abilities. We generated a mouse line in which the second exon of the tailless gene is flanked by loxP sites and crossed these animals with a transgenic line expressing the Cre recombinase in the neurogenic area of the developing brain, but not in the eye. The resulting animals have anatomically indistinguishable brains compared with tailless germline mutants, but are not blind. They are less anxious and much more aggressive than controls, like tailless germline mutants. In contrast to germline mutants, the conditional mutants are not impaired in fear conditioning. Furthermore, they show good performance in the Morris water-maze despite severely reduced hippocampal structures. Thus, the pathological aggressiveness and reduced anxiety found in tailless germline mutants are due to malformations caused by inactivation of the tailless gene in the brain, but the poor performance of tailless null mice in learning and memory paradigms is dependent on the associated blindness.

Thyrocyte-specific Gq/G11 deficiency impairs thyroid function and prevents goiter development.

The function of the adult thyroid is regulated by thyroid-stimulating hormone (TSH), which acts through a G protein-coupled receptor. Overactivation of the TSH receptor results in hyperthyroidism and goiter. The Gs-mediated stimulation of adenylyl cyclase-dependent cAMP formation has been regarded as the principal intracellular signaling mechanism mediating the action of TSH. Here we show that the Gq/G11-mediated signaling pathway plays an unexpected and essential role in the regulation of thyroid function. Mice lacking the alpha subunits of Gq and G11 specifically in thyroid epithelial cells showed severely reduced iodine organification and thyroid hormone secretion in response to TSH, and many developed hypothyroidism within months after birth. In addition, thyrocyte-specific Galphaq/Galpha11-deficient mice lacked the normal proliferative thyroid response to TSH or goitrogenic diet, indicating an essential role of this pathway in the adaptive growth of the thyroid gland. Our data suggest that Gq/G11 and their downstream effectors are promising targets to interfere with increased thyroid function and growth.

Conditional inactivation of the E-cadherin gene in thyroid follicular cells affects gland development but does not impair junction formation.

We have conditionally inactivated the E-cadherin gene in the thyroid follicular cells of mouse embryo to unravel its role in thyroid development. We used the Cre-loxP system in which the Cre-recombinase was expressed under the control of the tissue-specific thyroglobulin promoter that becomes active at embryonic d 15. At postnatal d 7, thyroid follicle lumens in the knockout mice were about 30% smaller with respect to control mice and had an irregular shape. E-cadherin was almost completely absent in thyrocytes, beta-catenin was significantly reduced, whereas no change in gamma-catenin was detected. alpha-Catenin was also reduced on the cell plasma membrane. Despite the dramatic loss of E-cadherin and beta-catenin, cell-cell junctions were not affected, the distribution of tight junction proteins was unaltered, and no increase of thyroglobulin circulating in the blood was observed. In addition, we found that other members of the cadherin family, the R-cadherin and the Ksp-cadherin, were expressed in thyrocytes and that their membrane distribution was not altered in the E-cadherin conditional knockout mouse. Our results indicate that E-cadherin has a role in the development of the thyroid gland and in the expression of beta-catenin, but it is not essential for the maintenance of follicular cell adhesion.

Estrogens and SERMs in coronary heart disease.

Estrogen combines beneficial and harmful actions by affecting many intracellular pathways in a large number of target organs related to the cardiovascular system. In observational studies and large outcome trials, an improvement of serum lipid profile and reduction of cardiovascular event rate were reported, whereas thrombembolic complications and stroke rate increased. Recognition of the diversity and tissue selectivity of estrogen's effects prompted the development of selective estrogen receptor modulators (SERMs), which were subsequently used to dissect the different mechanisms of action. SERMs are estrogen receptor (ER) ligands that exert partial agonist or antagonist actions on the ER in a tissue-, pathway- or isoform-specific manner. As ER ligands, they trigger a large variety of effects, including extranuclear ER actions, which can be further modulated by coactivators, corepressors and potential novel estrogen-binding proteins/receptors. Thus, SERMs can display tissue- or pathway-specific effects, or a combination of these. Pharmacological and clinical data are available for the classical SERM prototypes raloxifene and tamoxifene, as well as for new SERMs in different stages of development, isotype-specific agonists and pathway-selective ligands. These compounds exert many different effects, including vasodilatation in coronary arteries, altered responses to ischemic damage, hypertrophy of the myocardium, and improvement in serum cholesterol and lipid profile. The development of future SERMs will focus on different indications, including hormone therapy or cardiovascular disease. However, they all should antagonize estrogen action in female reproductive organs, yet protect from bone loss and not interfere with the beneficial effects of estrogen in the brain.

c-myc as a mediator of accelerated apoptosis and involution in mammary glands lacking Socs3.

Suppressor of cytokine signalling (SOCS) proteins are critical attenuators of cytokine-mediated signalling in diverse tissues. To determine the importance of Socs3 in mammary development, we generated mice in which Socs3 was deleted in mammary epithelial cells. No overt phenotype was evident during pregnancy and lactation, indicating that Socs3 is not a key physiological regulator of prolactin signalling. However, Socs3-deficient mammary glands exhibited a profound increase in epithelial apoptosis and tissue remodelling, resulting in precocious involution. This phenotype was accompanied by augmented Stat3 activation and a marked increase in the level of c-myc. Moreover, induction of c-myc before weaning using an inducible transgenic model recapitulated the Socs3 phenotype, and elevated expression of likely c-myc target genes, E2F-1, Bax and p53, was observed. Our data establish Socs3 as a critical attenuator of pro-apoptotic pathways that act in the developing mammary gland and provide evidence that c-myc regulates apoptosis during involution.

Definition of estrogen receptor pathway critical for estrogen positive feedback to gonadotropin-releasing hormone neurons and fertility.

The mechanisms through which estrogen regulates gonadotropin-releasing hormone (GnRH) neurons to control mammalian ovulation are unknown. We found that estrogen positive feedback to generate the preovulatory gonadotropin surge was normal in estrogen receptor beta knockout (ERbeta) mutant mice, but absent in ERalpha mutant mice. An ERalpha-selective compound was sufficient to generate positive feedback in wild-type mice. As GnRH neurons do not express ERalpha, estrogen positive feedback upon GnRH neurons must be indirect in nature. To establish the cell type responsible, we generated a neuron-specific ERalpha mutant mouse line. These mice failed to exhibit estrogen positive feedback, demonstrating that neurons expressing ERalpha are critical. We then used a GnRH neuron-specific Pseudorabies virus (PRV) tracing approach to show that the ERalpha-expressing neurons innervating GnRH neurons are located within rostral periventricular regions of the hypothalamus. These studies demonstrate that ovulation is driven by estrogen actions upon ERalpha-expressing neuronal afferents to GnRH neurons.

Protein-structure-based prediction of animal model suitability for pharmacodynamic studies of subtype-selective estrogens.

Subtype-selective estrogens are of increasing importance as tools used to unravel physiological roles of the estrogen receptors, ERalpha and ERbeta, in various species. Although human ERalpha and ERbeta differ by only two amino acids within the binding pockets, we and others recently succeeded in generating subtype-selective agonists. We have proposed that the selectivity of the steroidal compounds 16alpha-lactone-estradiol (16alpha-LE(2), hERalpha selective) and 8beta-vinyl-estradiol (8beta-VE(2), hERbeta selective) is based on the interaction of certain substituents of these compounds with essentially one amino acid in the respective ER binding pockets. For in vitro and ex vivo pharmacological experiments with these compounds we intended to use bovine tissues available from slaughterhouses in larger quantities. Using homology modeling techniques we determined that the amino acid conferring high hERbeta-selectivity to 8beta-VE(2) is not exchanged between human and bovine ERalpha and bovine ERbeta. Thus, we predicted our steroidal hERbeta-selective compound to exhibit only weak agonistic activity at bERbeta and that bovine tissue is therefore not suited for investigation of ERbeta functions. The situation is presumably identical for pig, sheep, and the common marmoset, whereas rats, mice, and rhesus macaques are appropriate animal models to study pharmacological effects of 8beta-VE(2) in vivo. This prediction was confirmed in transactivation studies assessing estradiol (E(2)) and the two subtype-selective ligands on bovine ERbeta and on a series of hERalpha and hERbeta with mutations in their respective ligand-binding pockets. We have shown that the detailed understanding of the interactions of a compound with its target protein enables the identification of relevant species for pharmacological studies.

Beta1 integrins regulate mammary gland proliferation and maintain the integrity of mammary alveoli.

Integrin-extracellular matrix interactions play important roles in the coordinated integration of external and internal cues that are essential for proper development. To study the role of beta1 integrin in the mammary gland, Itgbeta1(flox/flox) mice were crossed with WAPiCre transgenic mice, which led to specific ablation of beta1 integrin in luminal alveolar epithelial cells. In the beta1 integrin mutant mammary gland, individual alveoli were disorganized resulting from alterations in cell-basement membrane associations. Activity of focal adhesion kinase (FAK) was also decreased in mutant mammary glands. Luminal cell proliferation was strongly inhibited in beta1 integrin mutant glands, which correlated with a specific increase of p21 Cip1 expression. In a p21 Cip1 null background, there was a partial rescue of BrdU incorporation, providing in vivo evidence linking p21 Cip1 to the proliferative defect observed in beta1 integrin mutant glands. A connection between p21 Cip1 and beta1 integrin as well as FAK was also established in primary mammary cells. These results point to the essential role of beta1 integrin signaling in mammary epithelial cell proliferation.

Evaluation of steroid receptor function by gene targeting in mice.

Corticosteroid hormones regulate a variety of developmental, physiological and pathological processes via their cognate receptors, the glucocorticoid receptor (GR) and the mineralocorticoid receptor (MR). Using modern genetic technologies, including bacterial artificial chromosome-based transgenesis and conditional gene targeting, we have generated a panel of tissue-specific and function-selective mutations of the two corticosteroid hormone receptors in the mouse. These mouse models have allowed us to gain new insights into corticosteroid hormone signaling in vivo. By investigating a hepatocyte-specific GR mutation, it has been possible to define a novel biological action of GR, namely to function as a coactivator for Stat5-mediated gene transcription in the control of body growth. The investigation of brain-specific mutations have not only allowed us to better understand hypothalamo-pituitary-adrenal (HPA) axis regulation by glucocorticoids, but also to analyse corticosteroid action in various aspects of brain function like anxiety-related or addiction-related behaviour, and learning and memory. A function-selective mutation in the GR has allowed us to dissect different pathways in the gene expression regulation by this receptor, namely to separate DNA response element-binding dependent gene activation from response element-independent gene regulation via interference with other transcription factors. These different transcriptional activities of GR play an important role in glucocorticoid-mediated immunosuppression.

The epithelial glucocorticoid receptor is required for the normal timing of cell proliferation during mammary lobuloalveolar development but is dispensable for milk production.

Glucocorticoids have been shown to influence mammary gland function in vivo and to stimulate milk protein gene expression in vitro. Here, we describe the generation and analysis of a mouse model to study glucocorticoid receptor (GR, NR3C1) function in mammary epithelial cells. Using the Cre-loxP system, mutant mice were obtained in which the GR gene is specifically deleted in epithelial cells during lobuloalveolar development, leading to a complete loss of epithelial GR at the onset of lactation. Mice harboring the mammary-epithelial-specific GR mutation are able to nurse their litters until weaning. During pregnancy, however, GR deficiency delays lobuloalveolar development, leading to an incomplete epithelial penetration of the mammary fat pad that persists throughout lactation. We identified a reduced cell proliferation during lobuloalveolar development as reason for this delay. This reduction is compensated for by increased epithelial proliferation after parturition in the mutant glands. During lactation, GR-deficient mammary epithelium is capable of milk production and secretion. The expression of two milk proteins, namely whey acidic protein and beta-casein, during lactation was not critically affected in the absence of GR. We conclude that GR function is not essential for alveolar differentiation and milk production, but influences cell proliferation during lobuloalveolar development.

Targeting mammary epithelial cells using a bacterial artificial chromosome.

We describe the generation of transgenic mouse lines expressing Cre recombinase in epithelial cells of the lactating mammary gland. As an expression vector, we used a P1-derived bacterial artificial chromosome (PAC) which harbors the gene for the secretory milk protein, whey acidic protein (Wap). Using homologous recombination in E. coli, the PAC was modified to carry the improved coding sequence of Cre recombinase (iCre). Transgenic lines carrying the WAPiCre PAC express Cre recombinase efficiently in the majority of mammary epithelial cells upon lactation. Of only four transgenic lines produced, three express Cre recombinase to a high efficiency. LoxP-flanked DNA sequences are recombined in virtually all epithelial cells of WAPiCre transgenic mice at lactation day 3.