The study of changes in the morphology of neurons is important for understanding neural activity. In fact, changes in the dendritic arbor or the number of dendritic spines and axonal boutons, allow neurons to modify the strength of their synaptic input and output and to adapt to changing environments (for a review see Whitt et al., 2014).

Dendritic spines are membranous protrusions, which can establish excitatory or inhibitory synapses through different neurotransmitter receptors. Although they were thought to be a distinctive feature of principal neurons, different studies have shown that some interneuronal subpopulations also display these postsynaptic specializations (Freund and Buzsáki, 1996). In interneurons, however, the structure of the spines is slightly different: they lack the spine apparatus found in pyramidal neurons, several synapses are established per spine (in contrast with the one or two found in pyramidal neuron spines) and they are usually less numerous (Gulyás et al., 1992; Acsády et al., 1998). On the other hand, boutons are axonal thickenings, which contain and release the synaptic vesicles. Because of their roles as postsynaptic and presynaptic elements, spines and boutons have been found to be proper markers for neuronal input and output; therefore, increases in spine and axonal bouton density have been correlated to increases in neuronal activity (Engert and Bonhoeffer, 1999; Becker et al., 2008).

Excitatory neurons experience structural remodeling under different conditions, involving changes in the length and complexity of dendritic arbors, and in the density or morphology of their spines (Fu and Zuo, 2011) and axonal boutons (Florence et al., 1998; Colicos et al., 2001; Nikonenko et al., 2003). This has been shown in different conditions and disorders, including chronic stress and depression (McEwen, 1999; Qiao et al., 2016), obesity (Dingess et al., 2016), neurodevelopmental disorders (Glausier and Lewis, 2013; Flores et al., 2016) and after different pharmacological manipulations (Guirado et al., 2009; Yang et al., 2015; Castillo-Gómez et al., 2016b). By contrast, studies on the effects on interneuron morphology are scarcer, despite the important role of inhibitory networks in central nervous system physiology (Nacher et al., 2013). Most of these studies on interneuron plasticity are focused on basket interneurons and few have explored structural changes on dendrite-targeting interneurons, such as those expressing somatostatin. These interneurons are essential for the maturation of deep cortical circuits (Tuncdemir et al., 2016) and are important players in other stages of neurodevelopment, brain pathology and neuronal plasticity (for a review see Liguz-Lecznar et al., 2016). Only some recent studies have shown that these interneurons are able to undergo dendritic remodeling after chronic stress (Gilabert-Juan et al., 2011), antidepressant treatment (Guirado et al., 2014b), streptozotocin-diabetic challenge (Castillo-Gómez et al., 2015) or the depletion of plasticity related molecules (Guirado et al., 2014a; Castillo-Gómez et al., 2016a).

However, most of these structural analyses, as the majority of those in other interneuronal subtypes, have been performed on fixed tissue. Data acquired with this experimental approach is based on population analysis and does not allow for the longitudinal study of individual spines or boutons. Using such methods, neither transient nor homeostatic changes can be detected, i.e., changes that only alter the composition of synapses and leave the total number of spines or dendritic length intact. Structurally, these changes in network configurations are represented by the addition of one spine on one site, compensated by the pruning on another. The emergence of 2-photon microscopy and the chronic implantation of cranial windows have allowed for longitudinal analysis of individual elements of the neocortex after sensory deprivation, which have shown structural remodeling of excitatory (Hofer et al., 2009; Holtmaat and Svoboda, 2009; Cane et al., 2014), and inhibitory microcircuits (Chen et al., 2011a,b, 2012; Keck et al., 2011; van Versendaal et al., 2012; Chen and Nedivi, 2013). However, this technique is limited by the position of the cranial window and can only be used chronically, without brain damage, on studies of the neocortex. This inconvenience hinders the longitudinal study of the hippocampus, an important structure interconnected with the neocortex and responsible for spatial memory. The use of entorhino-hippocampal organotypic cultures, with their intrinsic limitations, helps to solve this problem and allows us to follow elements from a hippocampal neuron throughout an entire experiment. These cultures have been broadly used as an in vitro model of the rodent hippocampus (Stoppini et al., 1991), allowing the study of structural changes in real-time.

N-methyl-D-aspartate receptors (NMDARs) are a subtype of ionotropic glutamate receptors, expressed widely in both pyramidal neurons and interneurons (Collingridge et al., 1983; Nyíri et al., 2003; Alvarez et al., 2007; Oren et al., 2009). They play a key role in several events of central nervous system development, such as neuronal birth and migration (Komuro and Rakic, 1993). Antagonists to these glutamate receptors, such as MK-801, are known to interfere with the targeting and pruning of axons and the regulation of synaptogenesis during development (Cline and Constantine-Paton, 1990; Shatz, 1990; Butler et al., 1998). NMDAR antagonists also induce axonal sprouting during adulthood (Sutula et al., 1996; McKinney et al., 1999b) and are able to modulate some processes related to learning, such as LTP (Bailey et al., 1996). The blockade of these receptors does not appear to produce effects on the dendritic spine density of pyramidal neurons, neither in vivo (Woolley and McEwen, 1994) nor in vitro (McKinney et al., 1999b). However, this latter study showed the apparition of filopodia-like processes after chronic treatment with MK-801, resembling those in the developing hippocampus. Interestingly, NMDAR blockade with MK-801 increases the motility of dendritic spines of pyramidal neurons in hippocampal organotypic cultures (Alvarez et al., 2007). Regarding interneurons, little is known about the expression of NMDARs in these cells or about how the hypofunction of these receptors may affect their structure or physiology. Some studies in the prefrontal cortex have shown that MK-801 administration impairs the functional maturation of perisomatic inhibitory circuits expressing parvalbumin (Thomases et al., 2013) and affects differentially the physiology of these interneurons and pyramidal neurons (Wang and Gao, 2012). However, it is still unknown how this blockade may affect dendrite-targeting interneurons, such as those expressing somatostatin in the hippocampus. There are certain cell populations located in the stratum oriens sharing these characteristics, which are essential for the physiology of this region (Freund and Gulyás, 1997; Müller and Remy, 2014) and are known to present dynamic dendritic spines (Guirado et al., 2014a). Two of the most studied of these populations are the oriens-lacunosum moleculare (O-LM), and the hippocampo-septal (HS), interneurons (Jinno and Kosaka, 2002; Gulyás et al., 2003). The former are named after their microcircuitry: they receive their inputs in the stratum oriens from pyramidal neurons of the stratum pyramidale and reciprocally inhibit these cells and other interneurons, establishing these synapses in the stratum lacunosum moleculare (Müller and Remy, 2014). These cells are essential for the correct functioning of the hippocampus. In fact, they appear to mediate theta oscillations (Katona et al., 2014) and have been postulated to be essential for the establishment of spatial context-fear conditioning (Lovett-Barron et al., 2014; Müller and Remy, 2014).

In the present study, we aim to understand the impact of NMDAR manipulation on the structure and dynamics of somatostatin-expressing interneurons in the CA1 stratum oriens. In order to achieve our goal, we have used a transgenic mice strain in which these interneurons are constitutively labeled with enhanced green fluorescent protein (EGFP), which allows us to study their entire morphology (Oliva et al., 2000). After demonstrating the presence of NMDARs on these interneurons and their dendritic spines, we have acutely treated adult mice with the NMDAR antagonist MK-801 and have studied its effects on the structural remodeling of their dendritic spines and axonal en passant boutons (EPB). To have a behavioral readout of the efficacy of the treatment, we have measured locomotor activity, anxiety-related behavior and working memory. Finally, in order to study whether NMDAR manipulation modified spine dynamics in these cells, we have acutely blocked these receptors with MK-801 in entorhino-hippocampal organotypic cultures and have observed morphological changes in real-time.

The NMDARs are among the most studied receptors in the nervous system, because of their involvement in LTP and several developmental processes (Butler et al., 1998; McKinney et al., 1999b; Nacher and McEwen, 2006; Kehoe et al., 2014). These receptors also play an important role in the activity-dependent regulation of the morphology of the dendritic spines of excitatory neurons (Nikonenko et al., 2002; Ultanir et al., 2007). However, despite the abundant presence of NMDARs in interneurons (Nyíri et al., 2003), little is known about their role on the structural plasticity of these cells. Interneurons also change their axonal and dendritic morphology in response to different stimuli (Chen et al., 2011b; Gilabert-Juan et al., 2011; Nacher et al., 2013). Furthermore, some interneuron populations have spines in their dendrites and they are dynamic, as the ones present on pyramidal neurons (Keck et al., 2011; Guirado et al., 2014a). Only some studies have focused on the structural plasticity of interneurons. Dr. Nedivi’s laboratory, using cranial windows in transgenic mice with fluorescent interneurons, has shown the elongation/retraction of dendritic branch tips in vivo in different neocortical regions (for a review see Chen and Nedivi, 2013). There are scarce published studies on the turnover of interneuronal dendritic spines. In mice, these spines have a stability rate close to 98% in a 24 h time-lapse experiment (Keck et al., 2011). These results are considerable higher than the ones we describe in hippocampal interneurons in the control group, where the stability rate is around 70%. These differences may be due to a difference in the stability of different interneuronal subpopulations, the region and age studied, or, more likely, to a higher stability in vivo than in our organotypic cultures.

The use of NMDAR antagonists, such as MK-801 or ketamine, has increased exponentially our knowledge on the physiology of these receptors and their involvement in neurological and psychiatric disorders. They have been broadly used to model certain symptoms of schizophrenia (Gilmour et al., 2012), to produce neurotoxic insults (Beninger et al., 2002) or as a novel treatment for depression (Browne and Lucki, 2013; Yang et al., 2015). In order to obtain a behavioral readout of the effectiveness of NMDAR blockade induced by in vivo MK-801 administration, we used the hole-board apparatus and automatically tracked the mice to test for hyperlocomotion, anxiety-related behaviors, and effects on working memory. Our study agrees with the previously reported increase in locomotor activity in MK-801 treated mice (Zuo et al., 2006; Kalinichev et al., 2008). The working memory ratio showed a non-significant decrease in the MK-801-treated animals, which is similar to the absence of differences in head-dip counts reported by other labs (Haj-Mirzaian et al., 2015; Hirose et al., 2016).

There is some evidence of the impact of NMDAR antagonists on interneurons, particularly on those targeting the soma of pyramidal neurons (Romón et al., 2011; Rotaru et al., 2011). However, little is known about the effects that NMDAR blockade can exert on other types of interneurons, like those targeting the distal segments of the dendrites of pyramidal neurons. One of the best studied populations of these dendrite-targeting interneurons are the O-LM cells in the hippocampus. These interneurons are a specially interesting cell type to study structural plasticity, because they have dendritic spines (Sik et al., 1995; Katona et al., 1999; Guirado et al., 2014a), which mainly receive excitatory synapses from recurrent collaterals of local pyramidal cells (Blasco-Ibáñez and Freund, 1995). They are essential to the proper functioning of this region of the limbic system, since they appear to modulate CA3 and entorhinal inputs (Leão et al., 2012), mediate theta rhythm (Katona et al., 2014) and have been associated with spatial context-fear conditioning behavior (for a review see Müller and Remy, 2014). In addition to these O-LM cells, our study, using retrograde tracing has shown that a minority of the EGFP+ interneurons in the stratum oriens are HS cells, suggesting that NMDA receptors may also modulate the structure of these projection interneurons.

Interestingly, different studies have found that the dendritic complexity of somatostatin-expressing interneurons can change after a chronic stress paradigm (Gilabert-Juan et al., 2017) and that they particularly change their dendritic spine density after the chronic administration of fluoxetine (Guirado et al., 2014b).

Somatostatin-expressing interneurons in the stratum oriens express NMDARs in their somata and dendrites (Nyíri et al., 2003). In the present study, we increase this knowledge by showing NMDAR expression specifically in the dendritic spines of these interneurons. In hippocampal pyramidal neurons, in vitro studies have shown that NMDARs mediate spinogenesis during neurodevelopment (Kwon and Sabatini, 2011) and are responsible for spine structural remodeling (Matsuzaki et al., 2004; Lai and Ip, 2013). Therefore, it is not unreasonable to think that NMDARs could have a similar role in interneuronal spines and mediate these types of plasticity.

To know whether NMDARs could modulate interneuronal spine density in somatostatin-expressing interneurons of the hippocampus, we have acutely injected MK-801 to adult animals. Pyramidal neurons in the mPFC modify the density of their dendritic spines after NMDAR blockade with several drugs. Previous studies have shown that after NMDAR blockade with MK-801 in vivo, pyramidal spine density in the hippocampus remained unaltered (Woolley and McEwen, 1994; Han et al., 2013). In accordance with these results in excitatory neurons, we have found that the acute administration of MK-801 does not affect the density of dendritic spines of somatostatin-expressing interneurons in adult mice. However, other studies with NMDAR antagonists in pyramidal neurons have rendered positive results: studies using chronic treatments have described increases (Flores et al., 2007) or decreases (Velázquez-Zamora et al., 2011) in this parameter, and those using acute administration reported a rapid increase in spine density (Li et al., 2010; Liu et al., 2013; Phoumthipphavong et al., 2016). Discrepancies in the doses and duration of the treatment may account for these differential effects, as well as differences in the structure and physiology between the spines of pyramidal neurons and interneurons (Gulyás et al., 1992; Freund and Buzsáki, 1996; Acsády et al., 1998).

There is an apparent discrepancy between the results we obtained in vivo and those we found in organotypic cultures. This could be due to a higher spine stability in adult animals than in organotypic cultures (Guirado et al., 2014a). Moreover, the different results may be the consequence of a different subunit composition of their NMDARs. These receptors vary their composition during the second postnatal week, when there is a developmental switch from the subunit GluN2B to the subunit GluN2A, which have different kinetic traits (Erreger et al., 2005). Our hippocampal slices have been generated and studied during this time period, when their NMDAR still have the GluN2B subunit. The treatment of organotypic cultures with MK-801 induced significant changes in the dynamics and density of dendritic spines of somatostatin-expressing interneurons. Most of these cells are probably O-LM interneurons, since the type of organotypic culture used in our study (Stoppini et al., 1991) only preserves the perforant path and the local connections. Previous reports using this type of cultures have also shown effects of NMDAR antagonists on the structure of hippocampal pyramidal neurons: treatment with MK-801 for 7 days did not alter pyramidal spine density on CA1, but produced an increase in the number of filopodia-like processes, which resemble the immature spines observed during synaptogenesis in the developing hippocampus (McKinney et al., 1999a). Only a piece of previous work in primary cultures has shown that the chronic exposure to ketamine produces a retraction of the dendritic arbor of interneurons in cell cultures (Vutskits et al., 2007). Although the phenotype of these interneurons was not analyzed, these results are similar to those in the present study, which show a reduction in the density of dendritic spines and thus in the area susceptible for stablishing synaptic contacts.

Even though axonal remodeling has not been as broadly studied as that involving dendrites, some studies have also highlighted the effects of NMDAR antagonists on axonal structure. Only 3 days of MK-801 treatment are enough to produce axonal sprouting in the hippocampal Schaffer collaterals in vitro (McKinney et al., 1999b). We have observed a similar effect in adult animals on the axonal boutons of somatostatin-expressing interneurons. These collaterals in the stratum lacunosum moleculare are most likely emerging from O-LM cells, since they are the only somatostatin-expressing interneurons innervating this area (Freund and Buzsáki, 1996). Whether a similar effect on Schaffer collaterals occurs in vivo, still remains to be explored. However, it is interesting to note that a recent study has found that O-LM cells facilitate the transmission of intrahippocampal information through this input from CA3 and reduce the influence of the entorhinal afferents (Leão et al., 2012); the axonal terminal field of O-LM interneurons is located precisely in the same location of the entorhinal afferents to CA1 pyramidal neurons (Freund and Buzsáki, 1996).

Treatments with MK-801 have been extensively used to model the etiopathology of schizophrenia in rodents, because NMDAR antagonists can reproduce some symptoms of schizophrenia in normal individuals. In fact, there is a NMDA hypothesis of schizophrenia, which poses that alterations in the NMDARs, particularly during development, are involved in the etiopathology of this disorder (Adell et al., 2012; Lim et al., 2012). Since these receptors are expressed in certain interneuron populations, including somatostatin-expressing cells (Nyíri et al., 2003), it has already been speculated that a NMDAR hypofunction during development would inhibit interneurons, thus producing an overexcitation of the network. Different studies point to this hypofunction as a factor contributing to alterations of fast-spiking neurons in schizophrenia (Coyle, 2006; Lisman et al., 2008), but alterations of these receptors in other interneuronal populations could also result in an abnormal functioning or connectivity. Based in our present results and previous ones describing the structural plasticity of dendrite-targeting interneurons (Guirado et al., 2014a,b; Gilabert-Juan et al., 2017), it is reasonable to speculate that a hypofunction of NMDAR might result in changes in the postsynaptic (spines) and presynaptic (EPB) elements. Whether these changes occur in schizophrenic patients still remains to be determined. However, interestingly, in a recently developed double hit model of schizophrenia using the same strain of this study, we have observed a reduction of spine density in EGFP-labeled interneurons in the mPFC (Castillo-Gómez et al., 2017).

In conclusion, our results set a promising starting point to understand the dynamics of interneuronal spines, how this structural plasticity can be regulated by NMDARs and what are the mechanisms underlying it. With a better understanding of the structural remodeling of interneurons we may predict its impact on local neuronal networks and, in time, understand better the neurobiology of complex psychiatric disorders, such as schizophrenia or major depression, in which both NMDARs and inhibitory networks are affected.

EC-G, RG, JN and MP-R designed the study. MP-R, EC-G and RG performed the experiments. MP-R analyzed the data prepared the figures. MP-R, EC-G, JMB-I, CC, EV and JN wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.