International Union of Basic and Clinical Pharmacology CIV: The Neurobiology of Treatment-resistant Depression: From Antidepressant Classifications to Novel Pharmacological Targets

A. Definition of Treatment Resistance and Staging Models

In 1999, the GSRD implemented a staging method for treatment-resistant depression (Souery et al., 1999). According to their definition, the criteria for treatment resistance are fulfilled if a patient is resistant to at least two consecutive adequate antidepressant trials independently from the class of antidepressant (including augmentation and combination medications) administered. The different stages of treatment resistance correspond to the number of the following failed antidepressant trials (Souery et al., 1999).

Similarly, the European Medicines Agency (http://www.ema.europa.eu) defines treatment resistance as a nonresponse to at least two adequate antidepressant trials. In detail, the European Medicines Agency states: “TRD is considered, when treatment with at least two different antidepressant agents (of the same or a different class) prescribed in adequate dosages for adequate duration and adequate affirmation of treatment adherence showed lack of clinically meaningful improvement in the regulatory setting” (http://www.ema.europa.eu). Another staging model has been suggested by Thase and Rush (1997), considering a hierarchy of efficacy of different therapeutic strategies including also electroconvulsive therapy (ECT).

In the context of identifying treatment-resistant MDD conditions, it should be critically taken into account that some patients are considered to be treatment resistant even if they exhibit so-called “pseudoresistance” (i.e., a merely alleged resistance to the current antidepressant pharmacotherapy). Therefore, the debarment of “pseudoresistance” represents the first measure in case of insufficient response to the initial antidepressant monotherapy trial. Potential reasons for “pseudoresistance” can be, for instance, an inadequate dose and treatment duration of the antidepressant, insufficient plasma levels of the administered drugs, noncompliance of the patient with respect to medication intake, or relevant (nontreated) psychiatric and/or somatic comorbidities (Dold and Kasper, 2017) (Table 2).

B. Features Contributing to Treatment Resistance

In a comprehensive multicenter study of the GSRD, Souery et al. (2007) analyzed sociodemographic and clinical characteristics of 702 patients with MDD and found the following variables to be significantly associated with the presence of treatment resistance: comorbid anxiety disorders (panic disorder and social phobia), comorbid personality disorder, suicide risk, high depressive symptom severity, melancholic features, more than one previous hospitalization due to MDD, recurrent depressive episodes, nonresponse to the first administered antidepressant, and an age at onset of ≤18 (Table 3).

Applying machine learning algorithms to the above-mentioned GSRD patient sample, Kautzky et al. (2017) determined the timespan between first and last depressive episode, age at first antidepressant treatment, response to first antidepressant treatment, symptom severity, suicidality, melancholia, number of lifetime depressive episodes, patients’ admittance type, education, occupation, and comorbid diabetes, panic, and thyroid disorder to be the most useful predictors for treatment outcome.

In a recent further study of the GSRD, Balestri et al. (2016) investigated predictors for a very high degree of treatment resistance. Hereby, sociodemographic and clinical variables were examined in 98 patients with inadequate treatment response to at least three different antidepressants, including escitalopram and venlafaxine. In this clinical study, long duration and high severity of the current depressive episode, outpatient status, high suicidal risk, higher rate of the presence in the family history of psychiatric disorders, and the occurrence of adverse effects during the pharmacotherapy served as clinical predictors for severe treatment resistance. In another GSRD survey, Zaninotto et al. (2013) identified the following factors to be associated with treatment resistance: longer hospitalization over lifetime, longer duration of the current depressive episode, comorbid panic disorder, presence of melancholic and psychotic features, and suicide risk. However, it should be considered that this study (N = 699) was the first of all designed to determine differences between patients exhibiting psychotic or melancholic features (Zaninotto et al., 2013). Furthermore, Mandelli et al. (2016) investigated the impact of occupational levels on response patterns in 654 patients with MDD, whereby three occupational levels (high, middle, low) were compared with regard to the achievement of treatment response versus resistance. The analyses revealed a significant association between high occupational level and poorer treatment response in comparison with medium and low occupational levels (Mandelli et al., 2016). With respect to somatic comorbidities in MDD, no significant differences were found between responders and treatment-resistant patients in a sample of 702 patients with MDD (Amital et al., 2013). In terms of family history, Serretti et al. (2014) reported no statistically significant differences between patients with and without a family history of MDD when analyzing overall depressive symptoms. However, nonresponders with a family history of MDD showed higher core depressive symptoms compared with nonresponders without a family history of MDD (Serretti et al., 2014) (Table 3).

In addition to the aforementioned multicenter studies carried out by the GSRD, the finding of significantly poorer treatment outcome for patients with unipolar depression and concurrent anxiety was replicated in a study of the German Algorithm Project comprising 429 inpatients with MDD (Wiethoff et al., 2010). Similar findings were found in the large North American multicenter STAR*D (Sequenced Treatment Alternatives to Relieve Depression) study (Trivedi et al., 2006) investigating altogether 2876 outpatients with depression. Furthermore, largely corresponding findings stem from recent evidence derived from the STAR*D sample suggesting a final predictive model with accuracy about 70% and sensitivity about 88%, relating delayed remission rates to unemployment and severe baseline depression for instance (Falola et al., 2017).

A. Etiological Mechanisms of Major Depression

Depression is a multifactorial disease characterized by a heterogeneous group of symptoms associated with functional disability. A better understanding of the mechanisms and factors that contribute to disease onset is crucial not only for a better definition of different disease dimensions, but also to establish the molecular and functional alterations that may sustain specific symptoms. Such information may allow the identification of specific subgroups of patients with different sensitivity and responsiveness to pharmacological intervention. This possibility is in line with the NIMH Research Domain Criteria that represents a new way to classify mental disorders based on behavioral dimensions and neurobiological measures (Woody and Gibb, 2015). Indeed, the NIMH Research Domain Criteria initiative has the explicit goal of linking classification of psychopathology to the advances in genetics and neuroimaging across traditional diagnostic boundaries (Cuthbert and Insel, 2013). This may be further integrated with environmental and contextual influences to take into account the development or progression of a specific disease (Woody and Gibb, 2015). We postulate that different etiological mechanisms and their combination can lead to selected dysfunction that may be more or less sensitive to pharmacological intervention with classic drugs.

Despite strong evidence of heritability (Hyde et al., 2016), as described above, the efforts to identify the genetic underpinning of MDD have been largely unsuccessful, possibly due to disease heterogeneity and the absence of a biologic gold-standard for diagnosis. However, it is feasible to postulate that the major reason for the unsuccessfulness of genetic studies is represented by the strong contribution of environmental factors that interact and modulate genetic susceptibility (Klengel and Binder, 2013). With this respect, stress, in all its multiple forms, represents a key element for the risk to develop depression. There is, however, another important element that modulates the classic gene X environment interaction, which is time. Indeed, we know that genetic susceptibility factors may reshape developmental trajectories leading to altered ability to cope or respond to the environment within specific time frames (Babenko et al., 2015). Moreover, it is well known that exposure to stressful adverse events may have a different impact on brain function according to the timing of exposure, which can be clearly related to the maturation of different neuronal circuits that participate in stress response and that may serve to develop proper coping strategies.

It may be inferred that the functional outcome of stress exposure will not only depend upon the genetic background, but on the timing of stress exposure resulting in different effects in terms of circuits and brain structures affected, as well as molecular mechanisms that may sustain long-lasting modifications of brain function.

On a speculative basis, it may be possible to delineate specific vulnerability signatures as a result of stress exposure, based on the type and timing of the stressful experience, the genetic settings as well as other factors, including sex differences. These combinations may be associated with specific features (symptoms or dimension) of the disorder and may show a preferential responsiveness to a given therapeutic strategy.

A further complexity may be due to the fact that such “etiological” mechanisms may not necessarily produce a pathologic phenotype, but could represent a predisposing condition to develop MDD if reexposed to challenging precipitating traumatic experiences (Arloth et al., 2015; Pena et al., 2017). With this respect, epigenetic changes (see below) represent one important mechanism, through which a given system or cell population may keep the “memory” of early adversities, setting the stage for the onset of the disease later in life. The characterization of these epigenetic mechanisms is opening new possibilities to associate a given genotype with a specific pathologic phenotype.

A number of studies have been conducted all over the world with the purpose of reproducing selected etiological mechanisms to identify downstream changes relevant for specific behavioral and functional alterations associated with MDD. These studies have clearly defined some core mechanisms that are associated with MDD and that may contribute to different aspects of disease vulnerability and manifestation. The focus of these studies has now shifted from “classic” concepts, closely related to monoamine alterations, to more complex mechanisms, including reduced neuronal plasticity, synaptic dysfunction, enhanced inflammation, and altered hypothalamic-pituitary-adrenals (HPA) axis function and responsiveness. Although some of these aspects will be described in more detail in the next paragraphs, it is important to consider that there may be a close link between these alterations pointing to a cascade of events that may have a different origin based, for example, on etiological mechanisms, and will then propagate to affect global brain functioning. We believe that the ability to interfere with such network of changes represents a critical element for therapeutic response to pharmacological intervention, which should not only be able to restore normal function, but, more importantly, must promote resilience and reduce the long-term susceptibility to recurrent depressive episodes (relapse prevention).

In summary, the characterization of the etiologic mechanisms for MDD represents a key strategy to associate a given phenotype (for example, specific clinical features or specific symptoms) with a specific set of molecular and functional alterations. Although currently available drugs are known to interfere with synaptic mechanisms by blocking monoamine transporters or acting on different monoamine receptors, these agents may differ extensively with respect to their ability in modulating downstream mechanisms, which may be differentially affected in depressed subjects. This possibility can be particularly pertinent for treatment resistance that may be due to the inability of a given compound to effectively modulate one or more of these systems.

B. The Glutamatergic System

Depression has been historically defined a “monoaminergic disorder” according to the idea that the disease is due and sustained by a deficit of different monoamines, primarily noradrenaline and serotonin (Nestler et al., 2002). However, this view, while maintaining its importance, has been challenged by evidence and a number of questions that have brought about a revision of such a “hypothesis” that is now framed within more general concepts related to miswiring, deficits of neuronal plasticity, and cell-cell communication (Berton and Nestler, 2006). In this respect, evidence has pointed to glutamate as a crucial player in the etiology of depression and in treatment response (Mathews et al., 2012). Glutamate is the main excitatory neurotransmitter in the mammalian brain, and it plays a central role in memory processes and synaptic plasticity (Machado-Vieira et al., 2012) as well as in emotion regulation. Glutamate can affect neuronal activity and function in two different ways: 1) rapid actions, exerted via ligand-gated ion channels namely NMDA (N-methyl-d-aspartate), alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and kainate receptors and 2) slow modulatory actions, exerted via the eight G-protein coupled metabotropic receptors (mGluRs). When released in the synaptic cleft, its concentration is tightly regulated by glutamate transporters localized in both neurons and astrocytes (Danbolt, 2001). As mentioned above, preclinical and clinical evidence have associated glutamate with major depression. As an example, it has been hypothesized that a dysregulated excitatory (glutamatergic) neurotransmission in the ventral anterior (subgenual) cingulate cortex may result into a functional hypoactivity of the ascending monoamine systems (serotoninergic, noradrenergic, and dopaminergic) that contribute to the onset of affective and cognitive symptoms in MDD (Artigas, 2015). In particular, increased concentrations of glutamate and glutamine have been found in occipital cortex from MDD patients, whereas a decrease of the same neurotransmitters was detected in prefrontal regions (Hasler et al., 2007). An abnormal glutamate/glutamine/GABA cycling has been demonstrated in TRD patients (Price et al., 2009), where elevated Glx:GABA ratios were observed in occipital cortex, thus suggesting an impairment of glutamate-glutamine neuronal-glial cycling resulting in excessive buildup of extracellular glutamate and decreased glutamate release, leading to a reduction in cortical GABA (Sanacora et al., 2003; Price et al., 2009). A number of significant changes in the expression of glutamate receptors have been demonstrated in human postmortem studies (Gibbons et al., 2012). Altered glutamatergic function may contribute to reduced neuroplasticity and structural alterations that have been reported in the brains of subjects with depression as well as in animals exposed to chronic stress, which recapitulates a key etiological mechanism for major depression (Kim and Na, 2016).

Different studies in the last 20 years have demonstrated that a chronic treatment with “monoaminergic” antidepressants, such as tricyclics, strongly affects NMDA binding profiles and receptor function (Mjellem et al., 1993; Nowak et al., 1998). Moreover, the antidepressant-like effects of the SSRI escitalopram are prevented by NMDA receptor activation (Zomkowski et al., 2010). Coadministration of TCAs, such as imipramine, with amantadine (a noncompetitive NMDA receptor antagonist) reduced immobility time in the forced swim test (FST) in rats to a much greater extent than either treatment alone (Rogoz et al., 2002). Similar synergistic interactions have been observed between SSRIs and SNRIs and different uncompetitive NMDA receptor antagonists (Ates-Alagoz and Adejare, 2013), thus suggesting that NMDA receptor antagonists can enhance the preclinical efficacy of currently used monoaminergic antidepressants. On these bases, the glutamatergic system currently represents a field of major interest for drug discovery in TRD and for the development of new and more efficacious antidepressant drugs. Indeed, as described below (see Glutamatergic System), recent groundbreaking clinical studies have demonstrated that targeting the glutamatergic transmission is an effective and useful approach for treatment-resistant depression (Berman et al., 2000; Zarate et al., 2006; Sanacora and Schatzberg, 2015).

It has been hypothesized that by blocking NMDA receptors on GABAergic interneurons, ketamine causes a rapid, but transient, increase in extracellular glutamate in the prefrontal cortex (Duman, 2013). This process seems to involve spontaneous glutamate release, rather than typical evoked synaptic glutamate release. The consequent activation of AMPA receptors causes depolarization of postsynaptic neurons, leading to L-type voltage-gated calcium channels activation. As a consequence, BDNF released from vesicles activates the mammalian target of rapamycin, a signaling system that plays a central role in synaptic plasticity (Duman, 2013; Machado-Vieira et al., 2017) and that is known to be impaired in the prefrontal cortex of patients with TRD (Jernigan et al., 2011). Furthermore, blockade of NMDA receptors by ketamine can result in the inhibition of eukaryotic elongation factor 2 (eEF2) kinase, dephosphorylation of eEF2, which may lead to a desuppression of BDNF translation (Autry et al., 2011).

The role of mTOR as a target of ketamine has been observed also in humans, where ketamine is able to rescue mTOR signaling, as assessed in peripheral cells after acute administration of ketamine in MDD patients (Denk et al., 2011).

Recently, Zanos et al. (2016) claimed that the production of a distinct metabolite of ketamine [(2R,6R)-hydroxynorketamine (HNK)] is necessary and sufficient to produce the antidepressant effects of ketamine in mice through an NMDAR-independent pathway via sustained activation of AMPAR. However Suzuki et al. (2017) recently demonstrated that (2R,6R)-HNK inhibits synaptic NMDARs, triggering the same signaling pathways activated by ketamine (i.e., decreased eEF2 phosphorylation) and proposing that the sustained inhibition of NMDARs by (2R,6R)-HNK can explain the long-lasting antidepressant effects of ketamine in TRD, despite its relatively short half-life, which cannot be observed with other NMDAR antagonists. Nevertheless it should be noted that the potency of HNK for the NMDAR is much lower than ketamine and a functional inhibition of the NMDAR is observed only at concentrations that are higher than those reported as pharmacologically relevant for its antidepressant action in mice (Zanos et al., 2017). Moreover, ketamine and HNKs show comparable pharmacokinetic profiles, therefore ruling out the contribution of HNK for the protracted antidepressant activity of ketamine (Zanos et al., 2017). Future studies are needed to disentangle the complexity of the pharmacological and clinical activities of ketamine and its metabolites.

Ketamine has been promoting drug discovery processes in depression and TRD with the aim of developing ketamine-like molecules with the same clinical efficacy, but without psychotomimetic effects and an improved safety profile (Iadarola et al., 2015; Caraci et al., 2017b) (see Glutamatergic System).

C. Synaptic Plasticity and Neurotrophic Mechanisms

Antidepressant drugs have classically been associated with the ability to increase synaptic monoamines, mainly serotonin and noradrenaline, to restore the diminished levels that may contribute to specific symptoms of depression. However, as mentioned above, it is now well accepted that synaptic dysfunction may represent a core element for the pathologic phenotype (Calabrese et al., 2016). Indeed, structural alterations, including neuronal atrophy, reduced number of spines, and dendritic arborization have been consistently reported in the prefrontal cortex and hippocampus of depressed patients (Rajkowska et al., 1999; McEwen and Lasley, 2003; Stockmeier et al., 2004; Duman and Aghajanian, 2012). A link between structural alterations in depressed subjects and exposure to stress or traumatic experiences early in life has also been demonstrated (Duman and Duman, 2015), suggesting that such adverse events may lead to a protracted impairment of synaptic function and in cell-cell communication, thus leading to functional alterations of specific neuronal circuits (Negron-Oyarzo et al., 2016; Nephew et al., 2017). Interestingly, an inverse relationship between the total hippocampal volume and the duration of untreated depression has been demonstrated (Sheline et al., 1999). These alterations may result from toxic mechanisms, related to excessive glutamate, as well as to an overactivity of the glucocorticoid system (HPA dysfunction) (McEwen, 2001). Moreover, the extent of such structural alterations may be associated with and are paralleled by alterations of genes encoding proteins involved in synapse formation and function (Kang et al., 2012).

There are two important implications of these structural changes in relation to antidepressant response. On one end, regional brain volumes might be associated with rate and extent of clinical response to antidepressant medication (MacQueen et al., 2008). On the other end, smaller hippocampal volumes may predict lower response/remission rates in patients with depression treated with antidepressant drugs (Colle et al., 2016). A recent study has shown that patients with current depression had bilaterally reduced gray matter in the hippocampus compared with healthy control or untreated patients in stable remission. An increase in gray matter was observed in the hippocampus following treatment with citalopram in currently depressed patients. Moreover gray matter reduction in the hippocampus appears specific to the depressed state and can be considered a potential biomarker for a depressive episode (Arnone et al., 2013).

Although more information is needed to better define the relationship between such mechanisms, depressive state, and treatment responsiveness, these data suggest that the enhancement of both synaptic plasticity and synaptic strength may represent a critical aspect for the therapeutic effect of antidepressant drugs (Duman and Aghajanian, 2012). Although different mechanisms may contribute to structural and synaptic alterations of depressed patients, one class of proteins that play an important role in the maintenance of synaptic structure and function are neurotrophic factors, in particular the neurotrophin BDNF (Kuipers and Bramham, 2006; Greenberg et al., 2009; Park and Poo, 2013).

The role of BDNF in the pathophysiology of depression and in the mechanism of action of antidepressant drugs is well known (Altar, 1999; Duman, 2002; Duman and Monteggia, 2006; Bjorkholm and Monteggia, 2016; Calabrese et al., 2016; Cattaneo et al., 2016a). Several lines of evidence have shown that the expression of the neurotrophin is reduced in selected brain structures, as well as at the peripheral level, of subjects with depression (Shimizu et al., 2003; Karege et al., 2005; Thompson Ray et al., 2011; Reinhart et al., 2015). Similarly, reduced BDNF levels are found in chronically stressed rats and also in experimental models that show depressive-like behavioral alterations (Urani et al., 2005; Duman and Monteggia, 2006; Calabrese et al., 2009, 2015; Molteni et al., 2010a,b; Chourbaji et al., 2011; Luoni et al., 2014a,b; Berry et al., 2015). Interestingly, chronic treatment with different antidepressants can promote the expression of the neurotrophin and normalize its alterations in animal models (Calabrese et al., 2007, 2010; Molteni et al., 2009; Duric and Duman, 2013; Luoni et al., 2014b; Castren and Kojima, 2017). More importantly antidepressant treatments can normalize the alterations of peripheral BDNF levels observed in MDD patients (Cattaneo et al., 2010), an effect that correlates with symptomatology improvement (Sen et al., 2008; Cattaneo et al., 2013), suggesting a potential relationship between drug response and the ability to modulate neurotrophic mechanisms (Molendijk et al., 2011). Moreover, since it has been reported that pretreatment BDNF levels are directly correlated with antidepressant responses, the neurotrophin expression may also predict the response to antidepressants (Wolkowitz et al., 2011; Cattaneo et al., 2013).

Clinical studies have indeed reported alterations of BDNF system in TRD patients. For example, reduced BDNF gene expression was found in the blood of TRD patients (Hong et al., 2014). However, the effect of therapeutic intervention in TRD is not necessarily associated with the modulation of peripheral BDNF levels. Indeed, although it has been demonstrated that TRD patients who respond to ketamine show an elevation of serum BDNF (Duncan et al., 2013; Haile et al., 2014), other authors found that the effects of ketamine or ECT in TRD were not associated with changes in blood neurotrophin levels (Allen et al., 2015; Rapinesi et al., 2015).

All in all, these studies indicate that structural, synaptic alterations and changes in neuroplastic players, such as BDNF, play a crucial role in the pathophysiology of depression and may be highly relevant for TRD. Given that both synaptic dysfunction and deficits in BDNF system reflect compromised neuronal plasticity and consequently increased vulnerability to environmental risk factors, we might speculate that one potential therapeutic strategy for TRD should be to stimulate BDNF expression as well as synaptic mechanisms to promote resilience and counteract core alterations found in subjects with depression. In this respect, it must be kept in mind that the alterations of BDNF associated with depression are strictly dependent on the brain region considered. Indeed, although BDNF is downregulated at cortical and hippocampal level, opposite changes can be found in the mesolimbic system (Berton et al., 2006; Krishnan and Nestler, 2008; Wook Koo et al., 2016), suggesting that effective therapeutic intervention should be able to modulate the neurotrophin expression and function with anatomic selectivity. Recently, both tropomyosin receptor kinase B (TrkB) agonists and antagonists have shown a relevant antidepressant activity in animal models by rescuing BDNF signaling in the hippocampus and in the prefrontal cortex (TrkB agonists) or reducing its activity in the nucleus accumbens (TrkB antagonists) (Zhang et al., 2016). An alternative approach in this field might be the development of TrkB partial agonists able to rescue BDNF signaling in prefrontal cortex and hippocampus and to reduce the activity of this pathway in the nucleus accumbens, which may indeed lead to a significant antidepressant efficacy.

D. Hypothalamic-Pituitary-Adrenal Axis Dysfunction

Epidemiologic evidence supports a role for stress as a risk factor for depression (Cowen, 2010). Indeed, chronic exposure to stress can promote the development of major depression (Czeh and Lucassen, 2007; Pariante, 2017). Stressful events are known to activate the HPA axis, which finally stimulates the release of glucocorticoids from the adrenal cortex (de Kloet et al., 2005). Glucocorticoids are steroid hormones that readily cross the blood-brain barrier and bind to low-affinity glucocorticoid receptors (GR) and high-affinity mineralocorticoid receptors, exerting a physiologic negative feedback on HPA axis (de Kloet et al., 2005). It has been hypothesized that both high cortisol levels and the activation of immune system in MDD can be explained with the development of the so called “glucocorticoid receptor resistance” found in depressed patients. According to this idea, GR dysfunction may lead to an impaired function of the negative feedback, resulting in HPA axis hyperactivity and elevated cortisol levels (de Kloet et al., 2005; Maes et al., 2016; Pariante, 2017). GR resistance is particularly evident in patients with TRD (Bauer et al., 2003). Indeed, severe TRD is associated with an imbalance in the normal physiology of the HPA axis, with glucocorticoid receptor resistance combined with an increased mineralocorticoid receptor sensitivity (Juruena et al., 2013).

Different hypotheses have been made to explain the molecular links between HPA axis dysfunction, hypercortisolemia, and TRD pathogenesis. Cortisol increases the activity of tryptophan 2,3 dioxygenase (TDO), with an ensuing reduction in available tryptophan and a significant decrease of serotonin levels. Hypercortisolemia can also reduce neurogenesis in the hippocampal dentate gyrus (Krishnan and Nestler, 2008) and may lead to structural abnormalities, such as retraction of hippocampal apical dendrites (McLaughlin et al., 2007). Moreover, rats or mice exposed to social defeat show an activated response of the HPA axis (Keeney et al., 2006; Razzoli et al., 2009), which can be reversed by antidepressant treatments (Becker et al., 2008). Interestingly mice that are susceptible to social defeat stress show hypercortisolemia as well as significantly less GR protein expression and nuclear translocation in the hippocampus compared with resilient mice (Han et al., 2017). Accordingly, animals exposed to chronic mild stress (CMS) show increased expression of the chaperone protein FKBP5 as well as enhanced cytoplasmic levels of GR in hippocampus and prefrontal cortex (Guidotti et al., 2013).

Glucocorticoids can also contribute to MDD pathogenesis by reducing synaptic plasticity and increasing the vulnerability to neuronal death in the hippocampus (Yu et al., 2008). In particular glucocorticoids induce the expression of Dickkopf-1 (Dkk-1), an inhibitor of the canonical Wnt pathway, in hippocampal neurons and this may contribute to stress-induced structural alterations within the hippocampus (Matrisciano et al., 2011), which have also been consistently observed in MDD patients with a history of treatment resistance (Abdallah et al., 2015).

On these bases, preventing hypercortisolemia has been recently considered as a novel pharmacological strategy for MDD and, in particular, for TRD (Henter et al., 2017). Mifepristone, a glucocorticoid receptor antagonist, seems to be efficacious in the treatment of psychotic depression, a subtype of depression characterized by hypercortisolemia (Blasey et al., 2011), with a rapid improvement of depressive and psychotic symptoms (Belanoff et al., 2001, 2002). Moreover, one 6-week pilot study found that 600 mg/day of mifepristone improved depressive symptoms and cognition in patients with treatment-resistant bipolar depression, and this clinically relevant improvement was inversely associated with basal cortisol levels (Young et al., 2004). An alternative pharmacological approach to target hypercortisolemia is the use of metyrapone, an inhibitor of 11βhydroxylase, the enzyme that catalyzes the conversion of 11-deoxycortisol to cortisol (Sigalas et al., 2012). Preliminary clinical studies have shown the clinical efficacy of metyrapone in TRD. In particular a controlled, randomized, double-blind trial found that adjunctive metyrapone therapy to fluvoxamine was superior to placebo and accelerated the onset of antidepressant action (Jahn et al., 2004). A multicenter, placebo-controlled, randomized, phase 3 trial was conducted to evaluate the effects of metyrapone augmentation in patients with TRD, although the results of this study are not yet available (NCT01375920).

All in all, the ability to modulate HPA axis dysfunction as well as glucocorticoid receptor resistance represents an important aspect for the therapeutic action of antidepressant drugs and, together with other mechanisms, may be a key issue for treatment resistance (Pariante, 2017).

E. Immune System Dysregulation and Neuroinflammation

Over the last two decades several studies have demonstrated that inflammation and dysfunction of the immune system play a key role in the pathophysiology of major depression and may therefore contribute to treatment resistance (Caraci et al., 2010; Capuron and Miller, 2011; Maes et al., 2016; Remus and Dantzer, 2016; Bhattacharya and Drevets, 2017; Pariante, 2017). An altered activation of the immune system and the ensuing state of “peripheral and central inflammation” seem to be strictly correlated to HPA axis dysfunction observed in depressed patients (Remus and Dantzer, 2016; Pariante, 2017).

Depressed patients show higher levels of proinflammatory cytokines, such as interleukin-1 (IL-1), interleukin-2 (IL-2), interleukin-6 (IL-6), interleukin-8 (IL-8), interleukin-12 (IL-12), interferon-γ and tumor necrosis factor-α (TNF-α) (Dowlati et al., 2010; Capuron and Miller, 2011), as well as increased acute phase proteins, chemokines, and cellular adhesion molecules (Maes et al., 2016). In particular, recent meta-analyses found the most relevant longitudinal association between two inflammatory markers, namely C-reactive protein and IL-6, and depressive disorders, suggesting that indeed inflammation may contribute to the development of the disease (Valkanova et al., 2013; Smith et al., 2018). Moreover, elevated markers of microglial activation (measured by translocator protein binding in vivo with positron emission tomography) have been found in MDD patients (Setiawan et al., 2015). Interestingly, a positive correlation between the severity of the symptoms of depression and the increase in the inflammatory status has been demonstrated (Maes, 1999; Maes et al., 2016).

A number of preclinical studies have provided support for the role of immune/inflammatory dysfunction in depression (Remus and Dantzer, 2016; Leonard, 2018). As an example, lipopolysaccharide administration in rodents increases peripheral (and central) cytokines, such as IL-1 and TNF-α, leading to sickness behavior (reduced motor activity and a decrease in food and water intake) followed, 24 hours later, by depressive-like behavioral alterations (Ge et al., 2015; Remus and Dantzer, 2016; Sulakhiya et al., 2016). Accordingly, neuroinflammation and altered cytokine expression have been demonstrated in animal models of depression. Indeed, exposure to CMS leads to an increase of proinflammatory cytokines (IL-1β, TNF-α) and a decrease of anti-inflammatory cytokines (IL-10, IL-4, and TGF-β1) in different brain regions, as well as to enhanced markers of microglia activation (You et al., 2011; Hinwood et al., 2013; Rossetti et al., 2016). Interestingly the increase in inflammatory markers in dorsal hippocampus was inversely related to sucrose consumption, thus suggesting that the anhedonic-like phenotype in CMS rats may be linked to neuroinflammation (Rossetti et al., 2016).

Proinflammatory cytokines can interfere with many of the pathophysiological mechanisms relevant of depression (Maes et al., 2016). For example, interferon-γ and TNF-α induce the expression of the tryptophan-metabolizing enzyme indoleamine 2,3-dioxygenase (Campbell et al., 2014), the rate limiting step of the kynurenine pathway, and TDO (Leonard, 2007; Remus and Dantzer, 2016). Although it was proposed that such effects may lead to decreased serotonin levels (Remus and Dantzer, 2016), it was demonstrated that the reduction in peripheral blood TRP has no effects on CSF TRP concentrations (Raison et al., 2010). However, activation of indoleamine 2,3-dioxygenase and TDO leads to an increased production of the neurotoxins 3-hydroxykynurenine and quinolinic acid, which can contribute to the pathophysiology of MDD by activating the NMDA receptor (Myint and Kim, 2003; Leonard, 2007; Bay-Richter et al., 2015; Remus and Dantzer, 2016). Inflammatory cytokines strongly influence glutamate metabolism in astrocytes and microglia, and markers of inflammation correlate with dysfunction of glutamatergic system in the dorsal anterior cingulate cortex and are associated with anhedonia and psychomotor retardation. These studies suggest a strong neurobiological link between inflammation-induced depression and the dysfunction of glutamatergic system in TRD (Haroon and Miller, 2017).

It has been hypothesized that immune system activation and neuroinflammation can also lead to a deficit in dopaminergic mesolimbic pathway combined with a dysfunction in prefrontal glutamatergic system finally leading to the onset of anhedonia, loss of motivation, fatigue, psychomotor retardation, and cognitive deficits (Eisenberger et al., 2010; Leggio et al., 2013; Pan et al., 2017).

Neuroinflammation is also associated with a reduced response to the treatment with SSRIs (Kulmatycki and Jamali, 2006; Maes et al., 2016), and it may also account for the complex interaction of depression and cognitive deficits in older adults (Ownby, 2010).

Antidepressant drugs exert immune-regulatory effects, reducing the production of “peripheral” proinflammatory cytokines and stimulating the synthesis of anti-inflammatory cytokines such as IL-10 and TGF-β1 in patients with depression (Caraci et al., 2010; Maes et al., 2016). Recent studies suggest that antidepressant drugs may also exert direct anti-inflammatory effects on microglia (Tynan et al., 2012), which is known to be overactivated in MDD patients (Setiawan et al., 2015). Different antidepressants, including SSRIs, SNRIs, and the melatonergic drug agomelatine, are able to inhibit the production of different proinflammatory cytokines in vitro and in vivo (Tynan et al., 2012; Molteni et al., 2013; Ohgi et al., 2013). Moreover imipramine, agomelatine, and the novel antipsychotic drug lurasidone possess anti-inflammatory properties in the CMS model (Rossetti et al., 2016). Interestingly different preclinical and clinical studies have shown that ketamine is endowed with anti-inflammatory activity (De Kock et al., 2013) and reverses inflammation-induced depression in the lipopolysaccharide model by decreasing brain levels of inflammatory cytokines (Yang et al., 2013) or by blocking the effects of quinolinic acid, a downstream product of the kynurenine pathway, on NMDA receptors (Walker et al., 2013b). Furthermore, recent in vitro studies have shown that fluoxetine and venlafaxine increase the release of TGF-β1, an anti-inflammatory cytokine that is reduced in nonresponder MDD patients (Vollmar et al., 2008; Caraci et al., 2016).

However, although some markers of immune activation have been validated in TRD patients, some caution should be used when translating preclinical results with antidepressant drugs to the clinical setting. Nevertheless, immune parameters may predict treatment response in MDD patients. Indeed, nonresponder MDD patients show significant elevations in a variety of proinflammatory immunologic markers (Carvalho et al., 2013; Cattaneo et al., 2013; Kiraly et al., 2017), such as IL-1, macrophage inhibiting factor (MIF), and TNF-α (Cattaneo et al., 2013; Strawbridge et al., 2015). One of the most relevant studies in this field was conducted by Cattaneo et al. (2013) based on a large European Union-funded study, the Genome-based Therapeutic Drugs for Depression study (GENDEP). The authors provide evidence that MDD patients who do not respond to first- or second-generation antidepressants had higher baseline mRNA levels of IL-1, MIF, and TNF-α (Cattaneo et al., 2013), the levels of these three pro-inflammatory cytokines being able to predict about 50% of the variance in antidepressant response. The same group replicated these results in a second cohort of patients, showing that IL-1 and MIF mRNA levels can accurately predict antidepressant response in MDD patients with positive predictive values and specificity for nonresponders of 100% (Cattaneo et al., 2016b).

The link between inflammation and treatment resistance appears to be highly relevant for late-life depression (Alexopoulos and Morimoto, 2011). Geriatric depression and in particular “vascular depression” represents a specific clinical subtype of depression characterized by a low rate of response to “monominergic” antidepressants (Alexopoulos and Morimoto, 2011) and it is characterized by high levels of proinflammatory cytokines, such as IL-1β, IL-8, and IL-6 (Diniz et al., 2010; Taylor et al., 2013). Reduced levels of anti-inflammatory cytokines, such as IL-4, IL-10, and TGF-β1, have been found in the plasma of patients with depression (Maes, 1999; Myint et al., 2005; Musil et al., 2011; Rush et al., 2016) and can significantly contribute to treatment resistance in MDD (Musil et al., 2011). Interestingly responder and remitter patients with MDD had higher initial TGF-β1 levels at baseline compared with patients who did not respond to treatment (Musil et al., 2011). Patients with melancholic depression with a recent history of treatment resistance had significantly higher levels of the proinflammatory cytokine IL-6, and lower levels of the regulatory cytokine TGF-β1 than healthy controls (Rush et al., 2016). Deficit of TGF-β1 signaling is a common pathophysiological event both in depression and cognitive decline (Caraci et al., 2012), and the presence of cognitive symptoms in patients with depression might predict a low rate of response to current antidepressant drugs (Silverstein and Patel, 2011). We recently identified a key role for TGF-β1 in recognition memory formation demonstrating that this neurotrophic factor is essential for the transition from early to late long-term potentiation (Caraci et al., 2015). We hypothesize that a deficit of TGF-β1 may contribute to treatment resistance in elderly patients with MDD by increasing Aβ accumulation and the development of the so-called “amyloid-related depression,” a recently identified clinical phenotype in which the response to “monoaminergic” antidepressants is low (Li et al., 2017b).

F. Epigenetic Mechanisms

Epigenetic literally means “above genetics” and refers to changes in DNA structure without alterations of nucleotide sequence. The major epigenetic mechanisms are represented by DNA methylation, posttranscriptional histone marking, and by the control of mRNA processing and translation through noncoding RNAs (miRNAs) (Tsankova et al., 2007; Luoni and Riva, 2016).

Epigenetic processes are important mechanisms for experience-dependent changes in brain function and responsiveness and are considered key players for “long-term” maintenance of the effects produced by exposure to adverse events, particularly during early life stages. Epigenetic alterations may therefore represent a permanent scar of such events, thus contributing to the increased vulnerability for psychiatric disorders, such as major depression. This issue has been the topic of different excellent reviews that addressed several mechanisms linking environmental factors to mental illness through the complex interplay of epigenetic modifications (Tsankova et al., 2007; Graff and Tsai, 2013; Nestler, 2014; Han and Nestler, 2017). Moreover, although negative events occurring after brain maturation may exert limited and transient effects, insults experienced in a key period of development could reprogram the epigenome, and, being incorporated in the germ cells, the consequences of these events can also be transmitted to the progeny (Champagne, 2008; Bohacek et al., 2013; Babenko et al., 2015; Bale, 2015).

Epigenetic mechanisms appear to play an important role in major depression. On one end, the expression of different epigenetic regulators is altered in subjects with depression. As an example, histone deacetylases (HDACs) levels are significantly increased in peripheral blood cells (Iga et al., 2007; Hobara et al., 2010) and in the nucleus accumbens (Covington et al., 2009) of depressed patients, as well as in mice exposed to chronic stress (Renthal et al., 2007; Covington et al., 2009).

Furthermore, the dysregulation of important players in MDD appears to be sustained by epigenetic mechanisms. Indeed, in human brain the promoter of the GR gene is hypermethylated in men abused during childhood (McGowan et al., 2009) and in infants from mothers who self-reported depression during pregnancy (Braithwaite et al., 2015), as well as in a rat model of reduced maternal care (Weaver et al., 2004). Moreover, epigenetic mechanisms underlie the reduction of BDNF expression in an animal model of chronic stress (Tsankova et al., 2006). Similarly, a hypermethylation of BDNF promoter IV was observed in the Wernicke’s area of suicide completers (Keller et al., 2010). Finally, changes in DNA methylation and chromatin modification were reported in promoter regions of genes involved in protein synthesis (McGowan et al., 2008), polyamines (Fiori et al., 2012), and in neurotransmission (Abdolmaleky et al., 2006; Poulter et al., 2008).

While the investigation of selected genes does provide useful information with respect to the specific contribution of epigenetic mechanisms for functional alterations in MDD, genome-wide studies (epigenome-wide association studies) provide novel and important information on the epigenetic signatures that may be associated with a depressive phenotype (Sabunciyan et al., 2012; Labonte et al., 2013; Dempster et al., 2014).

It is therefore feasible to hypothesize that such mechanisms, which are not directly targeted by classic pharmacological intervention, may contribute to drug resistance. Accordingly, clinical evidence reports that HDAC2 and HDAC5 expression is upregulated in leukocytes of patients during the depressive phase but not in remission (Hobara et al., 2010) and that the DNA methylation profile of BDNF (exon I) allows for distinguishing between depressed and control patients (Fuchikami et al., 2011). On these bases, it is feasible to hypothesize that patients characterized by such abnormalities may require pharmacological interventions able to “act” at this level to show significant clinical improvement.

Epigenetic make up may also predict the response to antidepressant therapy. In particular, it has been reported that patients showing hypomethylation of the promoter of BDNF exon IV at plasma level are unlike to benefit from antidepressant therapy (Tadic et al., 2014), whereas responders to treatment showed a decrease of trimethylation of the histone 3 (Lopez et al., 2013). Accordingly, remitters/responders after ECT treatment showed a significantly lower methylation of exon I at the peripheral level compared with nonremitter/nonresponder subjects (Kleimann et al., 2015).

The relevance of these mechanisms in TRD is also suggested by the observation that drugs that may be effective in TRD act at epigenetic level (Tsankova et al., 2006; Vialou et al., 2013; Menke and Binder, 2014). For example, clinical doses of the valproic acid (VPA), a mood stabilizer used for the treatment of bipolar disorder, inhibit class I HDAC (Gottlicher et al., 2001; Kramer et al., 2003). In a small cohort of patients with severe TRD, antidepressant augmentation with VPA provides substantial clinical improvement and maintenance (Ghabrash et al., 2016). In rodents, chronic treatment with VPA alone significantly increases the gene expression of HDAC5, while combination with the antipsychotic lurasidone leads to a significant decrease of HDAC1 and two mRNA levels (Calabrese et al., 2013).

On another note, chronic treatment with suberoylanilide hydroxamic acid (also known as vorinostat), a class I and II HDAC inhibitor, partially rescues the molecular alterations and the depressive-like behavior of CRCT1^−/− (CREB-regulated transcription coactivator 1) mice, whereas conventional antidepressants, such as desipramine, do not show any effect (Meylan et al., 2016). Accordingly, infusion in the nucleus accumbens of suberoylanilide hydroxamic acid or another HDAC inhibitors (MS-275) rescues the depressive phenotype and the molecular alteration observed in the social defeat stress paradigm (Covington et al., 2009).

Moreover, HDAC inhibitors possess antidepressant properties in rodents as well as complementary procognitive actions also associated with neurodegenerative disease (Graff and Mansuy, 2008; Covington et al., 2009; Day and Sweatt, 2011; Lin et al., 2012; Yamawaki et al., 2012; Graff and Tsai, 2013; Noh and Seo, 2014). This is particularly important in depression, because cognitive dysfunctions have long been recognized as an intrinsic characteristic of major depressive disorder (Conradi et al., 2011; Millan et al., 2012) and represent the most common residual symptoms in partial responders (Gotlib and Joormann, 2010).

Interestingly, natural products with the ability to interfere with epigenetic mechanisms, such as folic acid and S-adenosylmethionin (fundamental for DNA formation and methylation), are lacking in depression and may be useful as adjunctive antidepressant therapy (Gilbody et al., 2007; Gomez-Pinilla, 2008; Sharma et al., 2017). Accordingly, L-acetylcarnitine, probably influencing the acetylation of H3K27 (Nasca et al., 2013), is able to normalize the phenotype of endogenously depressed Flinders sensitive line rats (Bigio et al., 2016).

However, it is worth mentioning that a major problem for any pharmacological intervention acting on epigenetic mechanisms is the lack of selectivity, and this may require additional studies before becoming an effective therapeutic strategy.

Although most of the results obtained at the peripheral level are in line with the changes observed in post mortem samples as well as in animal models, it remains to be established to what extent peripheral tissue/blood measures may be a proxy of brain changes in living humans.

On these bases, even if additional studies are needed, epigenetic mechanisms may represent an important target to develop new pharmacological treatments with potentially higher and more persistent efficacy in patients with TRD, although the lack of selectivity remains a major problem to be considered with such approach.

A. Dose Increase

With respect to dose increase strategies, a recent meta-analysis by Dold et al. (2017) found no evidence that nonresponders to an initial antidepressant trial benefit from a dose escalation of the same antidepressant drug. However, it should be taken into account that most of the included individual trials in this meta-analysis investigated high-dose treatment with SSRIs. This observation corresponds with a recent prospective study reporting that increasing escitalopram above the therapeutic range of serum escitalopram concentration seems to be useless with respect to improvement of antidepressant efficacy (Florio et al., 2017). It is noteworthy that these findings are in line with the current international treatment guidelines, suggesting that dose escalation cannot be currently regarded as a general evidence-based strategy for TRD (Bauer et al., 2017), although evidence based on open-label trials currently exists for a potential dose-response relationship of some tricyclic antidepressants (Hiemke et al., 2011) and the irreversible monoamine oxidase inhibitor tranylcypromine (Adli et al., 2008). Moreover, it should be considered that patients with polymorphisms in the cytochrome P450 enzyme system provoking an accelerated elimination of drugs (the so-called “ultra-rapid metabolizers”) probably require higher doses of the antidepressant to achieve treatment response if their plasma drug concentration is below the effective therapeutic range in a standard dose (Hiemke et al., 2011).

B. Switch of the Antidepressant Drug

According to available evidence and the current international treatment guidelines, a switch from one antidepressant drug to another new antidepressant after insufficient symptom improvement to the initial compound cannot generally be regarded as evidence-based treatment strategy despite potential advantages of continuing monotherapy (Bauer et al., 2017). In a recent meta-analysis of randomized controlled trials, switching was not superior to maintaining the pharmacotherapy with the initial antidepressant (Bschor et al., 2018). Following theoretical pharmacological considerations, it appears advantageous to choose preferably as a second, new antidepressant an agent with a different mechanism of action compared with the first administered compound. The World Federation of Societies of Biologic Psychiatry guidelines for instance advise explicitly that switching from an SSRI to venlafaxine or tranylcypromine appears justified within a treatment algorithm (Bauer et al., 2017). These recommendations are mainly based on the findings of a meta-analysis comparing a switch from an SSRI to either a second course of an SSRI or a switch to another class of antidepressants (Papakostas et al., 2008). As main result, slight but statistically significantly higher pooled remission rates for the latter strategy could be shown (28% for the across-class switch vs. 23.5% for the within-class switch) (Papakostas et al., 2008). In contrast to this meta-analytic finding, switching to a different subclass of an antidepressant (across-class switch) was not significantly superior to a within-class switch in a large European multicenter study (n = 340) when evaluating response and remission rates (Souery et al., 2011). In a recent meta-analysis of randomized controlled trials, switching was not superior to maintaining the pharmacotherapy with the initial antidepressant (Bschor et al., 2018). Accordingly, switching is generally not recommended as an appropriate treatment option for TRD and should only be employed in cases of absolutely no response or intolerable adverse effects (Bauer et al., 2017; Dold and Kasper, 2017). It is noteworthy that a careful selection of the initial antidepressant along with consistent therapy drug monitoring might minimize this risk (Serretti, 2018).

C. Antidepressant Combination Medication

Although antidepressant combination strategies are frequently used in the pharmacological management of TRD, the evidence for this measure is rather sparse, and meta-analytic findings on this topic were inconclusive (Rocha et al., 2012; Lopes Rocha et al., 2013). As the efficacy of this strategy depends primarily on the concurrently prescribed agents, treatment guidelines consistently advise establishing antidepressant combination preferably with reuptake inhibitors such as SSRIs or SNRIs on the one hand and inhibitors of presynaptic autoreceptors (e.g., mirtazapine) or serotonin antagonist and reuptake inhibitors (e.g., trazodone) on the other hand (Bauer et al., 2017). In this case, synergistic antidepressant effects can be anticipated due to the complementary mechanisms of action of these compounds (Moller et al., 2014). Furthermore, these drug combinations appear beneficial from a clinical viewpoint as presynaptic autoreceptor inhibitors are for instance—in contrast to SSRIs/SNRIs—characterized by meaningful sedating properties.

D. Augmentation Strategies

With regard to psychopharmacotherapeutic augmentation strategies in TRD, the most compelling evidence is available for SGAs and the mood stabilizer lithium (Dold and Kasper, 2017). The efficacy of SGA augmentation could be shown in a large number of randomized clinical trials and meta-analyses (Nelson and Papakostas, 2009; Zhou et al., 2015). For instance, in their meta-analysis of 16 placebo-controlled SGA augmentation trials (n = 3480), Nelson and Papakostas (2009) found significant superiority of adjunctive aripiprazole, olanzapine, quetiapine, and risperidone over placebo in response and remission rates. Moreover, some SGAs received the official approval as add-on medication after nonresponse to antidepressant monotherapy by regulatory authorities. For instance, quetiapine extended release (XR) is licensed in the United States and the European Union, aripiprazole in the United States, and olanzapine has the regulatory approval in the United States in combination with fluoxetine. Different mechanisms have been identified that can explain the efficacy of such combinations in clinical practice, such as 1) decreasing local inhibitory GABAergic tone in the dorsal raphe nucleus, through the antagonism of 5-HT₆ receptors by olanzapine, leading to a potentiation of SSRI’s activity (Asaoka et al., 2015) and 2) increasing hippocampal neurogenesis and preventing BDNF decrease in stressed rats (Xu et al., 2006).

Beside the SGAs, there is evidence for the efficacy of augmentation with lithium in treatment-resistant MDD (Crossley and Bauer, 2007; Bauer et al., 2014; Nelson et al., 2014), and, accordingly, treatment guidelines consistently recommend this strategy (Bauer et al., 2017). Nevertheless, adjunctive lithium was less frequently prescribed in pharmacoepidemiologic surveys than augmentation with SGAs (Dold et al., 2016). Probably, the use of lithium in the clinical practice is limited by the requirement of continuous plasma level measurements to ensure the achievement of the therapeutic window and due to the anticipation of adverse effects (Nierenberg et al., 2006).

Paralleled by the enhancement of available evidence especially for the efficacy of SGA augmentation, pharmacoepidemiologic studies consistently found a substantial rise of SGA prescription in MDD over the last years (Mohamed et al., 2009; Konstantinidis et al., 2012; Gerhard et al., 2014). For instance, a significant increase of the proportion of patients with MDD receiving SGAs from 12.8% in 2000 to 28.3% in 2007 was found in a pharmacovigilance program analyzing 1826 inpatients in German-speaking countries (Konstantinidis et al., 2012). From a clinical viewpoint, augmentation of antidepressants with antipsychotic drugs is especially recommended for patients with MDD exhibiting psychotic features, whereas lithium should be preferably considered in patients with MDD displaying high risk for suicidal behavior (Kessing et al., 2005; Cipriani et al., 2013; Bauer et al., 2017; Dold and Kasper, 2017).

A. Animal Models or Treatment-resistant Depression

The development of animal models in the field of psychiatry is very challenging because of the complex nature and the heterogeneity of these disorders. Therefore, although existing models may be able to reproduce specific features of the disease, they often show limited ability in mimicking the whole pathologic complexity, which is probably crucial for treatment resistance.

In a recent review, Willner and Belzung (2015) argued that among 18 potential models, the most promising are the Wistar Kyoto (WKY) and the congenital learned helplessness (cLH) rats, the CB1 (cannabinoid 1) receptor knockout, and organic cation transporter 2 null mutant mouse strains. The authors considered four criteria: increased stress responsiveness, decreased response to chronic antidepressant treatment, good response to novel antidepressant approaches, and the correspondence to a known clinical risk factor.

WKY rats were originally bred as normotensive control for the spontaneously hypertensive rat strain, but were later found to be characterized by heightened emotionality (Nam et al., 2014), enhanced response to repeated stress (Morilak et al., 2005), and increase propensity to develop learned helplessness (LH) (Belujon and Grace, 2014). Moreover, they are less responsive to antidepressants in the FST and less sensitive to serotonergic antidepressants (Lopez-Rubalcava and Lucki, 2000; Ivarsson et al., 2005). Notably, LH was normalized by ketamine in WKY rats, supporting the predictive validity of this model in the field of TRD (Belujon and Grace, 2014).

The cLH strain was generated by selecting animals that develop the most severe LH. After several breedings, cLH failed to escape without prior stress exposure (Vollmayr and Henn, 2001), they are anhedonic, and, different from noncongenital LH rats, their helpless behavior is not reversed by antidepressant treatment (Vollmayr et al., 2004; Sartorius et al., 2007; Vollmayr and Gass, 2013).

CB1 receptor knockout mice display increased CMS-induced anhedonia (Valverde and Torrens, 2012) and altered sensitivity to subchronic treatment with desipramine or paroxetine (Steiner et al., 2008), suggesting that they may represent a useful model for antidepressant resistance.

Organic cation transporter 2 null mutant mice show reduced levels of noradrenaline and serotonin, increased sensitivity to the effect of CMS (Courousse et al., 2015), and decreased response to chronic treatment with venlafaxine (Bacq et al., 2012).

A number of studies on treatment-resistant depression have also been carried out using other models, such as Flinders sensitive line (FSL) rats, CRCT1^−/− mice, and adrenocorticotropic hormone administration (Kitamura et al., 2002; Iwai et al., 2013). FSL rats are one of the most studied “depressive” rat strains (Wegener et al., 2012). Although they were originally generated for cholinergic supersensitivity, they show a range of depression-like features, such as elevated immobility in the forced swim test (Overstreet and Wegener, 2013) and increased anhedonia in the CMS (Pucilowski et al., 1993). Recently, it was demonstrated that exposure of FSL to time-dependent sensitization exacerbated their depressive-like behavior (Brand and Harvey, 2017). Interestingly, although chronic treatment with imipramine, venlafaxine, or ketamine was ineffective in normalizing their behavioral defects as monotherapy, the combination of imipramine with either venlafaxine or ketamine resulted in significant anti-immobility effects and enhanced coping behaviors (Brand and Harvey, 2017).

CRTC1-deficient mouse line exhibits reduced dopamine and serotonin turnover and decreased expression of genes involved in neuroplasticity in the prefrontal cortex. Moreover, CRTC1^−/− mice present several endophenotypes related to mood disorders, such as decreased sexual motivation, increased despair, anhedonia, and anxiety-like behavior (Breuillaud et al., 2012) as well as a blunted response to the antidepressants fluoxetine and imipramine in behavioral despair-related paradigms (Breuillaud et al., 2012; Meylan et al., 2016).

Chronic administration of adrenocorticotropic hormone in rats blocks the effects of antidepressants in the FST, alters monoamine response to stress, and downregulates HPA axis activity (Kitamura et al., 2002; Iwai et al., 2013; Walker et al., 2013a).

To summarize, unfortunately the animal models existing are few and, as mentioned above, show important limitations. Moreover, no animal model has been subjected to sequential application of different treatments a key requisite in human to be defined as TRD. Thus, there is still a large gap between clinical TRD and the possibility of effectively mimicking this condition in preclinical models.

B. Pharmacological Strategies in Treatment-resistant Depression