Pharmacogenomic Characterization in Bipolar Spectrum Disorders

The holistic approach of personalized medicine, merging clinical and molecular characteristics to tailor the diagnostic and therapeutic path to each individual, is steadily spreading in clinical practice. Psychiatric disorders represent one of the most difficult diagnostic challenges, given their frequent mixed nature and intrinsic variability, as in bipolar disorders and depression. Patients misdiagnosed as depressed are often initially prescribed serotonergic antidepressants, a treatment that can exacerbate a previously unrecognized bipolar condition. Thanks to the use of the patient’s genomic profile, it is possible to recognize such risk and at the same time characterize specific genetic assets specifically associated with bipolar spectrum disorder, as well as with the individual response to the various therapeutic options. This provides the basis for molecular diagnosis and the definition of pharmacogenomic profiles, thus guiding therapeutic choices and allowing a safer and more effective use of psychotropic drugs. Here, we report the pharmacogenomics state of the art in bipolar disorders and suggest an algorithm for therapeutic regimen choice.


Introduction
One important and recent development in the area of mood disorders is the recognition that many patients initially suspected suffering from major depressive disorder suffer instead from a form of bipolar disorder (BD) [1]. Symptomatic bipolar disorder patients are more frequently in the depressive condition rather than in hypomanic, manic, or mixed states [1][2][3]. As a consequence, such patients can be improperly diagnosed as suffering from major depression and prescribed antidepressants instead of the lithium, atypical antipsychotics, or antiepileptic-type mood stabilizers normally used to treat bipolar spectrum disorders [2,3]. Up to 50% of the patients once identified as suffering from unipolar depression are now recognized to be suffering from a bipolar spectrum disorder [1][2][3]. Antidepressant treatment of unrecognized bipolar patients can increase the cyclical nature of the mood disorder, mixed states, and conversion to hypomania and mania, and can also contribute to increase suicide in younger patients (<25 years) [4][5][6]. The World Mental Health Survey Initiative estimated a 12 month and total lifetime prevalence of 1.5% and 2.4%, respectively, for bipolar disorder I, bipolar disorder II, and subthreshold BD [5,7]. Most people are in their teens or early twenties when bipolar disorder symptoms begin. The diagnosis of one of the various bipolar spectrum disorders is very complex, especially during the early disease stages. It is estimated that only 20% of patients with one of the bipolar spectrum disorders associated with depressive episodes are diagnosed and treated correctly

Clinical Classification of Bipolar Spectrum Disorders
In general, bipolar disorders are mood disorders that, unlike depressive disorders, which are characterized by a single polarity, present manic or hypomanic episodes alternating with depressive episodes. They are generally divided into: • Bipolar I disorder (BDI): The characteristic that primarily characterizes this condition is alternating manic and depressive episodes. BDI is characterized by the appearance of one or more manic or mixed episodes (an overt phase of mania concomitant with a full-blown phase of depression) lasting at least a week. Patients with BDI can also experience episodes of major depression. • Bipolar II disorder (BDII): This condition is characterized by alternating depressive and hypomanic episodes. Type II bipolar disorder is a mental disease similar to type I bipolar disorder, with moods that cycle between highs and lows, although in BDII the "highs" never reach a complete mania state (hypomania). Subjects with bipolar II disorder suffer more frequently from depressive episodes than from hypomania. Given that hypomania can be confused with normal happiness or even normal functioning, bipolar II disorder can often be misdiagnosed as a unipolar depression.

Clinical Classification of Bipolar Spectrum Disorders
In general, bipolar disorders are mood disorders that, unlike depressive disorders, which are characterized by a single polarity, present manic or hypomanic episodes alternating with depressive episodes. They are generally divided into: • Bipolar I disorder (BDI): The characteristic that primarily characterizes this condition is alternating manic and depressive episodes. BDI is characterized by the appearance of one or more manic or mixed episodes (an overt phase of mania concomitant with a full-blown phase of depression) lasting at least a week. Patients with BDI can also experience episodes of major depression. • Bipolar II disorder (BDII): This condition is characterized by alternating depressive and hypomanic episodes. Type II bipolar disorder is a mental disease similar to type I bipolar disorder, with moods that cycle between highs and lows, although in BDII the "highs" never reach a complete mania state (hypomania). Subjects with bipolar II disorder suffer more frequently from depressive episodes than from hypomania. Given that hypomania can be confused with normal happiness or even normal functioning, bipolar II disorder can often be misdiagnosed as a unipolar depression.

Lithium
Lithium is used to decrease the recurrence of manic episodes, but it is also indicated for depressive episodes, although to a lesser extent [5]. Even though lithium has been used in the treatment of bipolar disorders for almost 60 years, its mechanism of action has not been fully clarified [23]. Some molecular steps have been implicated along the signal transduction cascade activated by neurotransmitter receptors, such as G proteins and phosphatidyl inositol. More recently, gene expression regulation of growth factors and neuronal plasticity has been associated with lithium activity via components of signal transduction, including protein kinase C and GSK3 [5,23,24]. Based on these and other putative molecular associations of the lithium pathways, several studies have been carried out in order to identify potential genetic lithium response predictors [25].

Antiepileptics as Mood Stabilizers
Based on the theory that the recurrence of manic episodes can expose the subject to further manic episodes (kindling), a logical parallelism has been drawn with epilepsy, where the appearance of repeated epileptic seizures exposes the subject to further seizures [5,26,27]. Several antiepileptics are often prescribed to treat bipolar disorders, some more effectively than others [26,27].

Valproic Acid
Valproate is recommended in long-term treatment to prevent mania recurrence in bipolar disorder. It is also used in the acute phase of mania, although its preventive activity has not been adequately clarified in this context [28]. Like lithium, valproic acid can be administered once a day with other mood stabilizers at doses that correspond to the lower limit of the therapeutic range, to enhance tolerability and therapeutic compliance [29]. As with all antiepileptics, its precise mechanism of action is uncertain. At least three hypotheses have been proposed: inhibition of the voltage-dependent sodium channel, enhancement of GABAergic neurotransmission, and regulation of downstream signal transduction cascades [28,29]. To explain its mood-stabilizing activity in mania, it has been hypothesized that valproate acts by reducing the excessive stimulation of neurotransmission, inhibiting the ionic flow through the voltage-dependent sodium channels (VSSC) [28]. No specific molecular site has been identified, but the drug may alter the phosphorylation of sodium channels, thus modifying their sensitivity. When less sodium enters the neuron, there is a reduction in glutamate release and therefore excitatory neurotransmission [28][29][30]. Others have hypothesized that valproate could enhance GABA activity by decreasing its re-uptake, increasing its release, or slowing its metabolic degradation. Although it remains unknown exactly how the enhancement of GABAergic tone is achieved, it is believed that this may explain the antimanic effect of valproic acid [28]. More recently, further mechanisms have been proposed for valproate that could explain its activity. Valproate can inhibit GSK3, but it can act on other molecular targets too; it can inhibit MARCKS (substrate of miranolated kinase C rich in alanine) as well as protein kinase C (PKC), and can activate various signals that promote long-term neuroprotection, such as BCL2, GAP43, ERK, and others [28][29][30].

Carbamazepine
Carbamazepine was the first antiepileptic with demonstrated effectiveness for mania in bipolar disorder [5,31,32], although it did not formally receive FDA approval. It is hypothesized that carbamazepine acts by blocking the voltage-dependent sodium channels (VSSC), possibly at the VSSC subunit level within the channel [33].

Lamotrigine
Lamotrigine acts as a mood stabilizer and it is used in prevention of depression and mania, although this use has not been approved by FDA in bipolar depression. Nonetheless, in many guidelines on treatment of bipolar depression, this drug is preferred to antidepressants as a first-line drug [5,34]. The reduction of excitatory-type glutamatergic neurotransmission may represent the specific mechanism of action of lamotrigine [35]. Some other antiepileptic drugs, like gabapentin, topiramate, oxcarbazepine/eslicarbazepine, and pregabalin, some calcium channel blockers of type L (e.g., dihydropyridine), and riluzole, are sometimes prescribed in "experimental" treatments for symptoms associated with bipolar disorder [5].

Atypical Antipsychotics
Atypical antipsychotics have been shown to be effective for the main non-psychotic symptoms of mania and for the prevention of recurrence of mania. Currently, they are the most effective therapeutic option for bipolar disorder, along with most antiepileptics and lithium [5,36]. The mechanism of action of atypical antipsychotics in bipolar disorder is not yet fully understood, but the prevailing hypothesis Pharmaceutics 2020, 12, 13 6 of 32 is that antagonism or partial agonism of D2 receptors could explain the reduction of manic psychotic symptoms. Moreover, the 5HT2A receptor (5HT2AR) antagonism and partial agonism of 5HT1A receptors could be responsible for the reduction of manic and non-psychotic depressive symptoms observed with some atypical antipsychotics. This may be achieved through the downregulation of the glutamatergic system affecting pyramidal neurons. Given that hyperactivity of the glutamatergic system, depending on the neuronal circuit involved, can be associated with both manic and depressive symptoms, these antipsychotics can be effective in reducing both types of symptoms. Other mechanisms have also been hypothesized to explain why some atypical antipsychotics improve the symptomatic picture of the depressive phase of bipolar disorder. All these mechanisms are based on the ability of some atypical antipsychotics to increase serotonin, dopamine, and norepinephrine levels, while reducing those of glutamate [37,38]. Atypical antipsychotics are indicated in schizophrenia, and most of them also in mania, but quetiapine is the only one approved for bipolar conditions, while lurasidone has only been tested [39,40].

Benzodiazepines
Although the use of these drugs as mood stabilizers has not been formally approved, benzodiazepines represent a treatment of considerable value, especially during emergencies. Their prompt administration can provide an immediate sedative effect and provide precious time, unlike mood stabilizers with a slower onset of action. Benzodiazepines are essential drugs for patients who suffer from intermittent episodes of agitation, insomnia, and incipient manic symptoms and require treatment as needed [41].

Antidepressants
Evidence is growing that antidepressants in these disorders not only do not work, but can even exacerbate the condition of some patients with bipolar disorder, leading to mania and hypomania states, destabilizing mood, and increasing cyclicality or even suicidality [2,3,42,43].
Evidence suggests that dysfunctions of glutamate neurotransmission may be implicated in various psychiatric conditions, including bipolar disorders [44,45]. The role of 5HT-mediated glutamatergic activation in BD could explain why the use of antidepressants can exacerbate manic symptoms in bipolar disorders. This hypothesis is still under study, but little doubt remains that antidepressants, tricyclic ones in particular, can trigger manic symptoms in subjects with bipolar spectrum disorders. 5HT2ARs are always postsynaptic and are located in many brain regions. In cortical neurons, they are coupled with Gαq/11 type G proteins. The latter activates membrane-bound phospholipase C beta, leading to cleavage of PIP2 into two messengers, IP3 and diacylglycerol (DAG). This stimulates protein kinase C (PKC), which in turn controls the function of the main glutamate transporter in CNS, Glutamate transporter-1 [46][47][48]. In particular, PKC-mediated phosphorylation induces GLT1 transporter downregulation/endocytosis, thus increasing glutamatergic intersynaptic activity [49][50][51][52].

Associations
In clinical practice, many subjects suffering from bipolar disorder need to be treated with more than one drug. Effective combinations include the association of valproate or lithium with an atypical antipsychotic. Evidence collected from clinical practice suggests more associations can be utilized, although such suggestions have not been adequately evaluated in controlled clinical trials. Examples include the combination of valproate and lithium, valproate and lamotrigine, lithium and lamotrigine, lithium with quetiapine, and lamotrigine with valproate and lithium. Expert opinions are very divergent when it comes to treating bipolar depression, particularly with antidepressants. Some believe that an antidepressant should not be given in any case, while others simply recommend caution when combining an antidepressant with a mood stabilizer.
In conclusion, current protocols and guidelines for acute mania recommend as first-line treatments lithium, quetiapine, divalproex, asenapine, aripiprazole, paliperidone, risperidone, and cariprazine in

Genetics of Bipolar Spectrum Disorders
Multiple genetic studies have pointed out that bipolar disorders (BPD) are often heritable conditions, with genetics accounting for 60-85% of the risk [54,55]. Studies indicate that the risk of recurrence of bipolar disorders in first-degree relatives is about 9%, almost 10-fold higher than in the general population [55,56]. Family studies have also indicated that bipolar I and II disorders have a genetic distinction; the risk of bipolar II disorder among relatives of patients with bipolar II disorder is greater than in relatives of patients with BDI [57][58][59].
Research on genes that might influence bipolar disorders has been hampered by the phenotypic and genetic complexity of the syndrome, a limited knowledge on its pathogenesis, and by the scarcity of animal models. Linkage studies have highlighted different chromosomal regions as carriers of meaningful genes, but with inconsistent results. It is now generally recognized that the genetic associations of bipolar disorder are linked to many different genes [60,61]. Thus, genetic research in the field has been focused on genome-wide association studies. The application of this approach to BD since 2007 has allowed the identification of a sizable number of candidate genes, which have since been associated with the disorder in various studies (including DAOA, BDNF, GRIK4, DISC1, TPH2, and SLC6A4) [62]. More recently, the first major BD genome-wide association study by the Psychiatric Genomics Consortium (PGC) Bipolar Disorder Working Group led to the identification of four significant loci at the genomic level. The study analyzed 7481 BD patients and compared them to 9250 controls. Three subsequent meta-analyses that included PGC BD data identified five more loci [63]. In one of the most important studies on the subject, a genome-wide association study was conducted with 20,352 cases compared to 31,358 controls. A total of 822 sentinel variants were followed up independently in 9412 cases versus 137,760 controls. As a result, 30 loci (with 20 new ones) achieved significant genomic association evidence and contained genes coding for synaptic components (ANK3, RIMS1) and transporters, neurotransmitters and ion channels (SLC4A1, CACNA1C, SCN2A, GRIN2A). Interestingly type I BD is genetically highly correlated with schizophrenia, while type II bipolar disorder correlates more with major depressive disorder [62,63]. In summary, the results of the broader genomic analysis on BD have revealed that these conditions have an extensive polygenic architecture, implicating in their etiology neurotransmitter and calcium channel functions, confirming that BD falls within a spectrum of highly related psychiatric disorders [55]. The detailed genetic dissection of the disorder, rationalized using a systems biology approach, could allow the identification of functional connections between the different molecular effectors identified, leading to the construction of a molecular network underlying the pathogenesis of the disease. This could also represent a valuable tool to guide, optimize, and personalize the therapeutic choice of molecular targets.

Pharmacogenomics of Bipolar Spectrum Disorder
The study of pharmacogenomics in the field of mental health is rapidly growing. Most data on genomics of bipolar spectrum disorders concentrate on the variability (response, side effects) of the pharmacological response when using atypical antipsychotics and/or antidepressants [36,64]. The genetic assets used in clinical settings are based mostly on the pharmacogenomic studies associating gene polymorphisms with treatment outcomes. In particular, genetic polymorphisms can affect pharmacodynamic and pharmacokinetic aspects of medications, thus influencing efficacy and susceptibility to side effects. With the increased accessibility of individual DNA chip analysis, WES (whole-exome sequencing,) and WGS (whole-genome sequencing) for diagnostic studies [65], pharmacogenomics is entering a phase of regular clinical use. The analysis of some key point mutations is actually required before specific drugs can be administered, in order to personalize the treatment according to the predicted efficacy or sensitivity to side effects [66]. Genetic variants to be analyzed for each class of drugs are selected according to available evidence-based medicine data and clinical validations, and several dedicated databases are available and freely accessible on the web today (e.g., pharmGKB [67], drugbank [68], genecards [69]).
Several factors can increase the possibility of a genetic diathesis in the treatment of bipolar spectrum disorders. Some meaningful SNPs are used in clinical settings to predict therapeutic response or potential toxicity. Polygenic determinants of drug effects have become more and more important in pharmacogenomics and are now used in clinical diagnostics to prevent adverse reactions to medications and to optimize therapy. Certain aspects of pharmacogenomic testing have entered the clinical routine. For example, FDA recommends testing for HLA-B*1502 when using carbamazepine. Carriers of this mutation are estimated to have a 10-fold higher risk for Stevens-Johnson syndrome when assuming the drug [70]. Today, the majority of tests available include both pharmacodynamic (PD) and pharmacokinetic (PK) genomic analytical panels.
Serotonergic pathways have been the focus of most pharmacogenetic studies on clinical response to psychotropic drugs in BD patients. Alterations of these pathways have been implicated in bipolar disorders. The serotonin transporter SLC6A4 gene has been extensively investigated. 5-HTTLPR is a functional polymorphism of SLC6A4, with a short (s) and a long (l) allele variant that have been suggested to be related to stress and psychiatric disorders. There have been reports associating the (s) allele with poor response to serotonergic drugs in BD. On the other hand, genetic studies on the (l) allele and SSRI response have not been able to clearly demonstrate a functional association between gene expression and pharmacological effect. For example, rs25531, a SNP located in the same SLC6A4 region as 5-HTTLPR, although associated with increased transporter expression, does not affect the response to SSRIs [75,76].
The HTR2A gene for 5HT-2A receptors (5HT2ARs) has been associated with antidepressant effect. Patients with specific polymorphisms of this gene (rs6313 and rs7997012) have been found to respond better to antidepressants. The same mutations have also been associated with an increased sensitivity of 5HT2ARs to serotonin [77,78]. 5HT2ARs hypersensitivity found in rs6313 and rs7997012 genetic mutations has been associated with the occurrence of pharmacologically induced dysphoric conditions in misdiagnosed bipolar spectrum disorders [15]. Several studies have shown an association between a good response to drugs that act on the serotonergic pathway in BD patients, and the rs6295 C/C genotype in the HTR1A gene [79][80][81]. The HTR2C gene for the 5HT-2C receptor has been shown to be associated with adverse drug reactions when using neuroleptic drugs. There is evidence indicating an association of specific polymorphisms like variant rs3813929 with a higher risk of extrapyramidal side effects. Another polymorphism, rs1414334, has been associated with higher risk of developing metabolic syndrome in subjects treated with olanzapine [82,83]. Dopamine 2 receptors have been associated with antipsychotic effect and are coded by the highly polymorphic DRD2 gene, located on chromosome 11q22. An extensively studied variant is -141C Ins/Del (rs1799732). When the Ins/Ins genotype is present, patients respond better to antipsychotic drugs than subjects carrying one or two copies of the Del allele. Subjects carrying the homozygous C allele with rs2514218 have been shown to respond better to antipsychotics than those homozygous for T allele, but they also present more side effects [84][85][86].
Pharmaceutics 2020, 12, 13 9 of 32 Little evidence is available on genes encoding for glutamate receptors. Research has been focused on the Gβ3 gene, with studies reporting an association between a good response to drugs that act on the serotoninergic pathway in BD patients, and the T/T genotype in the rs5443 polymorphism [79,87]. Table 1 shows some of the most important pharmacodynamic mutations that can modify the pharmacotoxicological outcomes of drugs in bipolar disorder.

Pharmacogenomics of Pharmacokinetic Pathways
CYP enzyme polymorphisms (cytochrome P450) may result in increased or decreased enzymatic activity. Such variations can thus determine a wide variety of drug metabolic patterns that, depending on the resulting metabolic activity, characterize the subject as a normal metabolizer, ultra-rapid metabolizer, intermediate metabolizer, or poor metabolizer. Poor metabolizers are subjects that have poor to no enzymatic activity, usually associated with having two copies of non-functioning alleles. Medications targeted by these enzymes are not metabolized effectively, thus increasing the risk of adverse drug reactions. In the case of pro-drugs, this condition can result in therapeutic failure, since the medication will not be transformed into the active form. Intermediate metabolizers have a slight functional impairment of drug metabolic activity due to at least one non-functioning allele. Normal metabolizers have a regular drug metabolic activity, while ultra-rapid metabolizers have enhanced enzyme activity that leads to an increased ability to metabolize drugs, decreasing the effectiveness of drugs, or increasing the risk of adverse drug reactions of prodrugs. For reference, the Clinical Pharmacogenetics Implementation Consortium (CPIC) published a review on pharmacogenomics nomenclature [161,162].
The FDA labeling (US Food and Drug Administration) of about 200 drugs specify their pharmacogenomic characterization. Neuropsychiatric medications represent one fourth of all these drugs. With pharmacogenomic information on the type of metabolizer, it is possible to choose the most effective drug for a patient, and identify the dose range and the administration strategy. Table 2 shows some of the most important pharmacokinetic mutations that can modify the pharmacotoxicological outcome of the drugs in bipolar disorder. Patients with the CYP2C19 * 1/* 1 genotype who are treated with citalopram or escitalopram may have an increased drug clearance/metabolism as compared to patients with CYP2C19 * 2, * 3, or * 4 allele and a decreased drug clearance/metabolism as compared to patients with CYP2C19 * 1/* 17 or * 17/* 17 genotype.

1A
Patients with the * 1 allele may have increased metabolism/clearance of risperidone as compared to patients with two reduced function alleles (* 10), one reduced function and one non-functional (* 4, * 5, or * 14) allele, or two non-functional alleles.

Conclusions
Bipolar spectrum disorder diagnosis is usually not based on the identification of pathogenetic mechanism, but on symptoms and signs, while the identification and association of traditional biological markers with bipolar disorder is still under investigation. The advent of genomics has allowed the identification of genetic assets associated with bipolar spectrum disorders, providing the basis for the identification of genetic risk factors and the definition of personalized pharmacotoxicological profiles. These can guide the initial therapeutic choice, or suggest corrections according to the individual's biological networks implicated in the disease pathways or in the relevant pharmacological aspects.
Even before they are diagnosed with a bipolar condition, patients usually claim a depressive symptomatology of some degree. Using the patient's molecular information can be helpful for the initial therapeutic orientation when a decision needs to be made on whether to use serotonergic antidepressants, given that mutations in their HTR2A gene would suggest a risk of manic state induction.
Once the bipolar condition is diagnosed, pharmacogenomic information can be used to guide the choice of the class of drugs to be used. In particular, the analysis of pharmacodynamically relevant SNPs can support the identification of the type of drugs with most chance of being effective and/or least likelihood of causing side effects. This information should be combined with a complete family history and medical information, as well as regular clinical risk factor profiling for bipolar disorder.
When the type of drug class to be used has been identified, pharmacogenomic information can help to drive the choice of the best molecule according to the individual pharmacokinetic asset, and at the same time indicate the dose and the therapeutic strategy that should be used to optimize effectiveness and minimize the risks of side effects.
A proposed algorithm is shown in Figure 2.
symptomatology of some degree. Using the patient's molecular information can be helpful for the initial therapeutic orientation when a decision needs to be made on whether to use serotonergic antidepressants, given that mutations in their HTR2A gene would suggest a risk of manic state induction.
Once the bipolar condition is diagnosed, pharmacogenomic information can be used to guide the choice of the class of drugs to be used. In particular, the analysis of pharmacodynamically relevant SNPs can support the identification of the type of drugs with most chance of being effective and/or least likelihood of causing side effects. This information should be combined with a complete family history and medical information, as well as regular clinical risk factor profiling for bipolar disorder.
When the type of drug class to be used has been identified, pharmacogenomic information can help to drive the choice of the best molecule according to the individual pharmacokinetic asset, and at the same time indicate the dose and the therapeutic strategy that should be used to optimize effectiveness and minimize the risks of side effects.
A proposed algorithm is shown in Figure 2. In summary, the genomic knowledge available today can support a personalized medicine approach to bipolar disorders, suggesting the most suitable pharmacological therapy for each patient. Together with the basic clinical information, a pharmacogenomic analysis should always be recommended to verify how one or more drugs might effectively act in a given subject, foreseeing the efficacy profile and safety of each drug, in order to increase therapeutic success and decrease unwanted adverse effects. Although more evidence is required before genomic data can be fully used to assist the clinician with a molecular diagnosis, when dealing with bipolar spectrum disorders, pharmacogenomics can provide orientation towards a safer use of drugs, particularly antidepressant In summary, the genomic knowledge available today can support a personalized medicine approach to bipolar disorders, suggesting the most suitable pharmacological therapy for each patient. Together with the basic clinical information, a pharmacogenomic analysis should always be recommended to verify how one or more drugs might effectively act in a given subject, foreseeing the efficacy profile and safety of each drug, in order to increase therapeutic success and decrease unwanted adverse effects. Although more evidence is required before genomic data can be fully used to assist the clinician with a molecular diagnosis, when dealing with bipolar spectrum disorders, pharmacogenomics can provide orientation towards a safer use of drugs, particularly antidepressant drugs, based on the polymorphisms on the HTR2A gene. Once a bipolar disorder is diagnosed, drugs can be chosen based on the molecular targets identified by the pharmacodynamic SNPs, optimizing the doses and the pharmacological combinations based on the pharmacokinetic SNPs.
Funding: This research received no external funding.

Conflicts of Interest:
The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.