Membrane-Derived Phospholipids Control Synaptic Neurotransmission and Plasticity

Synaptic communication is a dynamic process that is key to the regulation of neuronal excitability and information processing in the brain. To date, however, the molecular signals controlling synaptic dynamics have been poorly understood. Membrane-derived bioactive phospholipids are potential candidates to control short-term tuning of synaptic signaling, a plastic event essential for information processing at both the cellular and neuronal network levels in the brain. Here, we showed that phospholipids affect excitatory and inhibitory neurotransmission by different degrees, loci, and mechanisms of action. Signaling triggered by lysophosphatidic acid (LPA) evoked rapid and reversible depression of excitatory and inhibitory postsynaptic currents. At excitatory synapses, LPA-induced depression depended on LPA1/Gαi/o-protein/phospholipase C/myosin light chain kinase cascade at the presynaptic site. LPA increased myosin light chain phosphorylation, which is known to trigger actomyosin contraction, and reduced the number of synaptic vesicles docked to active zones in excitatory boutons. At inhibitory synapses, postsynaptic LPA signaling led to dephosphorylation, and internalization of the GABAAγ2 subunit through the LPA1/Gα12/13-protein/RhoA/Rho kinase/calcineurin pathway. However, LPA-induced depression of GABAergic transmission was correlated with an endocytosis-independent reduction of GABAA receptors, possibly by GABAAγ2 dephosphorylation and subsequent increased lateral diffusion. Furthermore, endogenous LPA signaling, mainly via LPA1, mediated activity-dependent inhibitory depression in a model of experimental synaptic plasticity. Finally, LPA signaling, most likely restraining the excitatory drive incoming to motoneurons, regulated performance of motor output commands, a basic brain processing task. We propose that lysophospholipids serve as potential local messengers that tune synaptic strength to precedent activity of the neuron.


Introduction
Activity-dependent plasticity of neuronal networks refers to the adaptive changes in their properties in response to external and internal stimuli. In a prominent form of central nervous system (CNS) plasticity, synaptic strength results in an increase (potentiation) or decrease (depression) of transmission efficacy, depending on the neuron's precedent activity (activitydependent synaptic plasticity). Short-lived processes that modify synaptic strength occur in practically all types of synapses [1], and short-term synaptic plasticity is essential in regulating neuronal excitability and is central to information processing at both cellular and neuronal network levels [2]. Homeostatic adjustment of synaptic weights counteracts neuronal rate disturbances that affect self-tuning neuronal activity within a dynamic range via Ca 2+ -dependent sensors [3]. The number of receptors in the surface membrane and at synaptic sites, and the size of the readily releasable pool (RRP) of synaptic vesicles (SVs), are important determinants of synaptic strength, short-term plasticity, and intersynaptic crosstalk [4][5][6][7][8]. Unmasking the feedback mechanisms that are believed to sense neuron activity and adjust synaptic strength (i.e., activity-dependent, coupled messenger synthesis and/or release) would help to explain how circuits adapt during synaptic homeostasis, experience-dependent plasticity, and/or synaptic dysfunctions that underlie cognitive decline in many neurological diseases.
The ligand-gated ionotropic channels-A-type GABA A receptors (GABA A Rs) and AMPAtype glutamate receptors (AMPARs)-mediate fast synaptic transmission at the vast majority of inhibitory and excitatory synapses, respectively, in the mammalian brain [4,5,9]. Cell surface stability of receptors is further regulated by post-translational phosphorylation, palmitoylation, and/or ubiquitination. In particular, AMPAR and GABA A R phosphorylation modulates the receptor's biophysical properties and membrane trafficking. Hence, the coordinated activity of kinases and phosphatases plays a pivotal role in controlling synaptic strength and neuronal excitability. Key residues within the intracellular domains of diverse AMPAR and GABA A R subunits are targeted by a number of kinases, including protein kinases A and C, calcium/ calmodulin-dependent kinase II, and tyrosine kinases of the Src family. Generally, phosphorylation stabilizes the receptor on the surface and, conversely, dephosphorylation appears to be important for receptor endocytosis [4,9].
Lysophosphatidic acid (LPA) is a strong candidate to function as a local messenger that rapidly affects synaptic strength. A membrane-derived bioactive phospholipid that affects all biological systems, LPA is an important intermediary in lipid metabolism and has a vital role in de novo biosynthesis of membrane phospholipids [10]. The nervous system is markedly modulated by LPA signaling. LPA, autotaxin (the main LPA-synthesizing enzyme), and many subtypes of LPA-specific G-protein-coupled receptors (LPA [1][2][3][4][5][6] ) are enriched in the brain [10][11][12]. Downstream signaling cascades mediating LPA signaling include mitogen-activated protein kinase (MAPK) activation, adenylyl cyclase inhibition or activation, phospholipase C (PLC) activation/Ca 2+ mobilization and/or protein kinase C (PKC) activation, arachidonic acid release, Akt/PKB activation, and the activation of small GTPase RhoA and subsequent Rho kinase (ROCK) stimulation [10]. Many subtypes of LPA receptors (LPARs) are expressed in the brain; in particular, LPA 1 is highly expressed and is the most prevalent receptor subtype in both the embryonic and adult brains [13][14][15]. Accordingly, LPA targets all CNS cell types to modulate developmental processes including neurogenesis, migration, differentiation, and morphological and functional changes [10]. However, little is known about how LPA signaling influences neuron physiology and neuronal connectivity or integrates incoming synaptic drive. Presynaptic LPA 2 at glutamatergic synapses mediates neuronal network hyperexcitability in an epileptic mouse model [16]. In addition, LPA 1 -deficient mice manifest alterations in managing diverse neurotransmitters [17][18][19][20]. Endogenous ROCK activity, an intracellular partner in LPA signaling, is necessary to maintain afferent AMPAergic and GABA A ergic synaptic strength in motoneurons [8]. As a conventional link in synaptic plasticity, activity-dependent LPA production occurs downstream of noxious activation of glutamate receptors in models of neuropathic pain [21]. However, whether LPA signaling is actually able to modulate synaptic strength and mediate activity-dependent synaptic plasticity remains unresolved.
The aim of this study was to investigate whether LPA regulates synaptic strength and plasticity of motoneuron excitatory and inhibitory synapses. Here, we show that LPA-mainly via LPA 1 -induced rapid and reversible depression in synaptic strength (short-term depression [STD]), and operated as an autocrine messenger mediating activity-dependent STD at inhibitory synapses. At glutamatergic synapses, presynaptic LPA signaling reduced the size of the RRP of SVs. At GABAergic synapses, postsynaptic LPA action mediated dephosphorylation and endocytosis-dependent internalization of the GABA Aγ2 subunit. Strikingly, LPA signaling regulated the performance of motor output commands in vivo. Therefore, LPA seems to have important implications for synaptic plasticity, pathology, and information processing in the brain.

Results
The hypoglossal motor system was used as an experimental model to test the hypothesis that LPA regulates synaptic function. Hypoglossal motoneurons (HMNs) are arranged in the hypoglossal nucleus (HN) at the dorsal medulla, being easily accessible for functional studies in animal models. In vitro, AMPAR-and GABA A R-mediated neurotransmission incoming to HMNs can be feasibly isolated and are well characterized [8]. From an experimental point of view, a considerable advantage of this system is that the inspiratory-related afferent activity in HMNs, almost exclusively mediated by AMPAergic signaling, persists even in the in vivo decerebrated preparation [22,23]. Interestingly, HMNs underpin essential motor commands for normal suckling behavior in the neonate, a vital activity altered in LPA 1 -deficient mice [24]. In addition, lpa 1-6 mRNAs are expressed by HMNs in the adult mouse (Allen Mouse Brain Atlas, http://mouse.brain-map.org/; [25]). Altogether, previous findings point to this motor system as a suitable model to investigate the role of LPA in the control of excitatory and inhibitory synaptic neurotransmission at the CNS.

Anatomical Support for a Role of Lysophospholipids at Excitatory and Inhibitory Synapses
To explore a possible role of LPA in shaping the normal motor output of the HN, it was necessary to determine the predominant isotype of its main target receptors expressed in this motor nucleus. Assessment of the expression levels of mRNAs for LPA 1-6 receptors in microdissected HN from neonatal (P7) rats revealed that lpa 1 mRNA was 1.5 to 12.5 times more abundant than lpa 2-6 transcripts ( Fig 1A). Subsequently, confocal analysis of double immunolabeled HN from P7 pups showed LPA 1 -immunoreactive (ir) puncta, patches, and fiber-like structures Expression pattern of LPA 1 in the HN supports a role for lysophospholipids in the control of synaptic neurotransmission. (A) Expression levels of mRNA for indicated LPARs obtained by qRT-PCR of microdissected HNs from neonatal rats (P7) relative to the housekeeping GAPDH. Values were normalized taking mean value for lpa 1 as 1. *p < 0.05, one-way ANOVA on Ranks relative to lpa 2-6 . Plot data can be found in S1 Data. (B-K) Multiple immunolabeling confocal images of the HN from P7 rats showing a close relationship between LPA 1 -ir patches and structures expressing the nonphosphorylated form of neurofilament H (SMI32), a motoneuron marker (B-F), VGLUT2-(G, H, open arrows), and/or VGAT-ir (I-K) inputs. The 18.8 ± 2.1% (n = 35 HMNs) of VGLUT2-ir and 11.9 ± 1.6% (n = 51 HMNs) of VGAT-ir inputs apposed to HMN somata colocalized with LPA 1 -ir patches. Note that in 3-D reconstructions (E, F, and Hb), LPA 1 staining colocalizes and borders SMI32-ir somata and dendrites (E, F, arrows). The yzand xz-planes also confirm colocalization of LPA 1 -ir with excitatory (Ha, open arrows) and inhibitory (K) inputs. Finally, VGLUT2-(Hb) and VGAT-ir (I) inputs appeared apposed on LPA 1 -containing SMI32-ir dendrites. The xzand yz-planes are located as indicated by the white dashed lines. Scale bars: B, 50 μm; G, 25 μm; C, D, and J, 10 μm; E and I, 5 μm; and F, H, and K, 2 μm. colocalizing with SMI32-positive HMN perikarya and dendrite-like structures (Fig 1B-1D). Three-dimensional reconstructions agreed with a cytoplasmic and membrane localization of LPA 1 in perikarya and main dendritic branches of HMNs (Fig 1E and 1F). Triple immunofluorescence for LPA 1 , SMI32, and the vesicular glutamate (VGLUT2) or GABA (VGAT) transporters as synaptic markers confirmed that LPA 1 -ir puncta were colocalizing with excitatory (VGLUT2-ir) or inhibitory (VGAT-ir) presynaptic structures (Fig 1G, 1H, 1J, and 1K). Both excitatory and inhibitory inputs were also found apposed to SMI32-ir neuropil or somata coexpressing LPA 1 (Fig 1H and 1I). Although LPA 1 expression in other neural cell types is not excluded, this expression pattern supports pre and/or postsynaptic roles of LPA 1 at the main excitatory and inhibitory inputs on HMNs, suggesting a potential contribution of LPA to motoneuron physiology.

LPA Induces STD of Excitatory and Inhibitory Inputs in a Dose-Dependent Manner
Next, we investigated the functional effects of LPA on glutamatergic and GABAergic synaptic currents by whole-cell patch-clamp recordings of HMNs (slices from P6-P9 rats). Electrical stimulation of the ventrolateral reticular formation (VLRF) evoked postsynaptic currents (ePSCs) in HMNs (Fig 2A). The AMPAR-or GABA A R-mediated components of ePSCs (excitatory [eEPSCs AMPA ] or inhibitory postsynaptic currents [eIPSCs GABAA ], respectively) were isolated and recorded as described in S1 Text.
The two major species of LPA (approximately 70%) found in the brain [26], monounsaturated (18:1, or LPA) and saturated (18:0, or s-LPA), were used in this study. While LPA activates LPA 1-3 , s-LPA has high affinity for LPA 1/2 , but is a comparatively poor agonist against LPA 3 [27]. Unless stated otherwise, LPA was used at a similar concentration (2.5 μM) to that found in serum (1-5 μM) [28]. In general, unsaturated LPAs are more potent than s-LPA in activating LPARs and inducing biological activities [29]. Accordingly, a higher concentration was used for s-LPA (40 μM) than for LPA (2.5 μM) to achieve a similar effect on neurotransmission. Both phospholipids, added for 10 min to the bath solution, strongly attenuated the amplitude of eEPSCs AMPA and eIPSCs GABAA (Fig 2B). The effects were reversed after 10 min of washing. Thus, LPA modulated rapidly and reversibly the strength of AMPAR-and GABA A Rmediated synaptic transmission in motoneurons.
The tested dose (2.5 μM) of LPA had a proportionately higher effect on inhibitory than on excitatory inputs (Fig 2B and 2C). Further, differential sensitivity to LPA was studied by applying various concentrations, ranging from 1 nM to 20 μM. After subtracting vehicle-induced changes (S1 Text), an effect on both currents was detectable at concentrations as low as 10 nM and increased with LPA concentration to a similar maximum reduction in both currents (approximately 70%) at 10-20 μM (Fig 2C). Dose-response relationships were well fitted (p < 0.001; r 2 > 0.99) by biphasic (two slopes) five-parameter logistic equations, suggesting that LPA affects synaptic neurotransmission by multiple mechanisms. It remains to be determined whether this is the consequence of the recruitment of diverse isoreceptors and/or downstream signaling pathways. In any case, from the nanomolar to first-order micromolar range, LPA diminished inhibitory inputs (IC 50 = 1.0 ± 0.17 μM) in greater proportions (p < 0.001, Kolmogorov-Smirnov test) than excitatory ones (IC 50 = 1.8 ± 0.08 μM), but at higher concentrations, LPA affected both synaptic systems similarly (Fig 2C).

LPA Operates Presynaptically at Excitatory Inputs
As in our previously published study [8], a combined electrophysiological analysis was performed to identify the LPA synaptic site of action. LPA signaling on AMPAR-mediated transmission is likely not attributable to changes in postsynaptic sensitivity to glutamate. LPA did not alter the amplitude in both the miniature quantal EPSCs AMPA (mEPSCs AMPA ) and Fig 2. LPA induces STD at excitatory and inhibitory synapses in a dose-dependent manner. (A) Schematic diagram of the in vitro experimental model used to analyze the effects of LPA on synaptic transmission incoming to HMNs. Whole-cell patch-clamp recordings (Rec) were obtained from somata of HMNs in neonatal brain stem slices. Experiments were performed as in our previous published study [8]. ePSCs were evoked by electrical stimulation (Stim) of the VLRF. A micropipette (arrow) near to a HMN before patch performance is illustrated in the inset. (B) Top, schematic representation of the experimental protocol carried out to study drug effects on ePSCs. Motoneurons were initially allowed to stabilize (Stabil.) with normal artificial cerebrospinal fluid (aCSF) to obtain baseline control recordings. Slices were then superfused for 10 min with aCSF alone (vehicle) or with LPA or s-LPA (drug) before current responses were acquired again. Finally, a last round of acquisition was taken after a 10 min washout with drug-free aCSF. Bottom, examples of eEPSCs AMPA (left panels) and eIPSCs GABAA (middle panels) recorded from HMNs before and following exposure to LPA (2.5 μM) or s-LPA (40 μM) for 10 min and after washing. Right panels, reduction in the eEPSCs AMPA (blue; n > 10 HMNs per drug) or eIPSCs GABAA (green; n = 4 HMNs per drug) amplitude for LPA-or s-LPA-treated groups compared to their respective pretreatment (before) periods. eEPSCs AMPA or eIPSCs GABAA were pharmacologically isolated in the presence of 1 μM strychnine hydrochloride, 30 μM d-tubocurarine, 50 μM (DL)-APV, and 10 μM bicuculline methochloride or NBQX (20 μM), respectively, continuously applied to the bath perfusion. (C) Left panel, examples of eEPSCs AMPA (top) and eIPSCs GABAA (bottom) recorded from two HMNs before and during exposure to the indicated LPA concentrations. This experimental design was carried out in three HMNs per each synaptic category resulting in dosedependent attenuation of the ePSCs similar to those presented in the plot. Right panel, reduction in eEPSCs AMPA and eIPSCs GABAA amplitude induced by LPA at the indicated concentrations relative to control (before) condition. Each HMN was exposed to a single dose of LPA. Data for each drug concentration were averaged from at least three independent experiments. Mean changes in amplitude obtained after 10 min incubation with vehicle (aCSF, S1 Text) were subtracted from alterations induced by each tested LPA concentration. The study included only those motoneurons that recovered after washing to at least the percentage of change obtained by vehicle perfusion. n ! 4 HMNs per each concentration and synaptic signaling system. *p < 0.05, one-way RM-ANOVA relative to ePSCs recorded before lysophospholipid incubation. #p < 0.05, one-way ANOVA on Ranks relative to reduction in amplitude of eEPSCs AMPA measured after the same LPA concentration. Plots data can be found in S1 Data. postsynaptic currents evoked by exogenous glutamate pulses (S1 Text; S1 Fig). For that reason, we sought evidence for a presynaptic mechanism by recording spontaneous AMPAergic synaptic currents under facilitated spontaneous glutamate release (sEPSCs AMPA ). In this condition, synaptic activity was a mixture of action potential-dependent and -independent events. After LPA treatment, the sEPSCs AMPA amplitude, but not frequency (10.8 ± 1.0 Hz, p = 0.761), reversibly decreased to a value similar to that recorded for mEPSCs AMPA in control condition (before: 36.0 ± 3.8 pA; LPA: 24.0 ± 2.0 pA; Fig 3A-3C). This agrees with a LPA-induced full inhibition of action potential-dependent events.
In addition, we evaluated eEPSCs AMPA facilitation using paired-pulse and repetitive afferent stimulation protocols as in our previously published study [8]. Under repetitive stimulation, a change in the amount of facilitation is considered to be attributable to a presynaptic change in the release probability of neurotransmitter quanta [1]. In the control condition, paired-pulse stimulation displayed a strong facilitation of eEPSCs AMPA over the entire range of interstimulus intervals tested, but this was more pronounced at shorter interstimulus intervals ( At this point in our study, the attenuation of eEPSCs AMPA induced by LPA was related to a reduction in the glutamate release probability, which is believed to be determined by the number of fusion-competent SVs or the size of the RRP of SVs [6,7]. This idea was further strengthened by a subsequent analysis of eEPSCs AMPA amplitude using the minimal stimulation paradigm, designed to stimulate only one fiber and a single or small number of release sites. As in our previous study [8], the intensity of the stimulation was set to elicit eEPSCs AMPA with 30% to 40% failure ( Fig 3E). In this context, LPA treatment evoked a significant reduction of the mean amplitude of eEPSCs AMPA elicited by minimal stimulation and an enhancement of the eEPSCs AMPA failure rate ( Fig 3E; S4 Fig). The presynaptic action of LPA on glutamatergic inputs is further supported because LPA 1 -ir puncta colocalize with Munc13-1, a presynaptic active zone (a.z.) marker [30], in VGLUT2-containing boutons (Fig 3F and 3G). The LPA 1 association with a region of the presynaptic membrane compromised in the fusion of SVs supports that LPA signaling has a direct relationship with the machinery involved in the regulation of neurotransmitter release.

LPA Modulates Excitatory Inputs via LPA 1 /G α i/o -PLC
The qRT-PCR and immunohistochemical studies, together with additional pharmacological tests (S1 Text: S5 Fig; S6 Fig), robustly point to LPA 1 as a pivotal LPAR affecting glutamatergic synapses. In this context, injection of a small interfering RNA (siRNA) against lpa 1 (siRNA lpa1 ; 2 μg/2 μl) into the fourth ventricle efficiently reduced LPA 1 expression in the brain stem ( Fig  4A; S1 Text; S7 Fig). siRNA lpa1 robustly diminished, but did not fully avoid, (s-)LPA-induced alterations on eEPSCs AMPA amplitude and PPR relative to the administration of control noninterfering siRNA (cRNA; 2 μg/2 μl) or vehicle (RNase-free phosphate buffered saline; 2 μl) ( Fig  4A-4D). Whether the remaining response of eEPSCs AMPA to (s-)LPA could be due to residual LPA 1 expression or to recruitment of compensatory mechanisms-e.g., via up-regulated LPA 3 in response to LPA 1 knockdown-remains to be elucidated. LPA 1 couples with and activates three G proteins: G α12/13 , G αi/o , and G αq/11 [10]. Previous findings [8] and pharmacological data (S1 Text) did not support G α12/13 involvement.  Fig 2B. (B) Mean sEPSCs AMPA amplitude for LPAtreated group (2.5 μM) compared to their respective pretreatment (before) and washout periods (n = 4 HMNs). (C) Normalized cumulative probability distributions of sEPSCs AMPA amplitude for each condition. Bin width: 2 pA. Note that the cumulative distribution of sEPSCs AMPA amplitude shifted to the left (p < 0.001; Kolmogorov-Smirnov test). (D) Top, eEPSCs AMPA recorded in a HMN at the indicated conditions in response to paired-pulse stimulation of VLRF. The rightmost trace shows the superimposition of the responses scaled to the peak of the first eEPSCs AMPA . Bottom, comparison of PPR measured at specified interpulse intervals for HMNs recorded at the indicated conditions (n = 4 HMNs). Paired-pulse ratio (PPR) was obtained from the amplitude of the first and second eEPSCs AMPA by the formula eEPSCs AMPA 2/eEPSCs AMPA 1. The stimulus intensity was adjusted so that the eEPSC AMPA 1 was approximately 50% of maximal amplitude, then maintained constant throughout the recording period.

LPA Signaling Reduces the Size of RRP of SVs via MLCK in Excitatory Boutons
LPA induces smooth muscle contraction in a PLC-dependent, ROCK-independent manner that involves myosin light chain (MLC) phosphorylation by MLC kinase (MLCK) [31]. These findings point to MLCK as a potential kinase mediating the presynaptic action of LPA on excitatory neurotransmission. Accordingly, LPA increased the p-MLC:MLC ratio in the HN relative to aCSF-incubated brain stem slices, which was fully prevented by coincubation with the specific MLCK inhibitor ML-7 (Fig 5A and 5B). In concordance, though ML-7 per se did not alter the amplitude of eEPSCs AMPA , as we also recently reported [8], it fully suppressed LPA-induced alterations on eEPSCs AMPA amplitude and PPR (Fig 5C-5F). This further supports MLCK as a main molecular substrate activated by LPA signaling within excitatory presynaptic terminals.
MLC phosphorylation stimulates actomyosin interactions [32], and presynaptic Ca 2+ concentration regulates MLCK activity and modulates the RRP size in the calyx of the Held synapse [33]. Therefore, LPA signaling, through its modulatory control on MLCK and the actomyosin cytoskeleton, might regulate clustering and spatial distribution of SVs within excitatory (S-type, spherical SVs-containing) boutons (S1 Text). Electron microscopy analysis, performed as in our previous study [8], showed that, in a MLCK-dependent way, LPA noticeably reduced the number of SVs near the a.z. in S-type boutons attached to HMNs, compared to control conditions (Fig 5G-5L; S1 Text). In addition, LPA induced a drop (−20.2 ± 6.3%) in the SV population morphologically docked to (i.e., in contact with) the a.z., which corresponds to the release-ready neurotransmitter quanta [34] that was prevented by coaddition of ML-7 ( Fig 5M and 5N). These outcomes robustly support that LPA signaling regulates the size of the RRP of SVs in S-type boutons by a MLCK-dependent mechanism.
Together, these data strongly suggest that the depression of synaptic strength induced by LPA treatment is dependent on a reduction in the probability of release from excitatory glutamatergic terminals. This effect is attributable, at least in part, to a reduction in the size of the eEPSCs AMPA evoked at 0.2 Hz by minimal stimulation of VLRF in HMN before and after treatment with LPA. Characteristically, the intensity of the stimulation was set to elicit eEPSCs AMPA with 30% to 40% failure at the control (before) condition. Bottom, mean eEPSCs AMPA amplitude (left) and failure rate (right) at indicated conditions (n = 4 HMNs). Experiments and analysis described in A-E have been performed as in our previously published study [8]. *p < 0.05, one-way (B, E) or two-way (D) RM-ANOVA relative to control (before) condition. (F) Confocal images of the HN obtained from P7 rats processed by triple immunolabeling for LPA 1 , VGLUT2 and the presynaptic active zone (a.z.) marker Munc13-1. Note triple colocalizations within the boxed areas. (G) 3-D reconstruction showing LPA 1 expression in the presynaptic a.z. of a glutamatergic bouton. Note that LPA-ir colocalizes with Munc13-1 and with a VGLUT2-ir SVs pool in the three planes. The xzand yz-planes are located as indicated by the white dashed lines. Scale bars: F, 5 μm; G, 2 μm. Plots data can be found in S1 Data.   showing the average ratio of pMLC to total MLC densitometric intensity for the control and treated slices. Ratio values were normalized relative to the control group. Columns represent the average of at least three independent experiments. *p < 0.05, one-way ANOVA on Ranks relative to control condition. (C) eEPSCs AMPA recorded from two HMNs in normal aCSF and after 10 min bath perfusion with the indicated combination of drugs. (D) Average eEPSC AMPA amplitude for the ML-7 (n = 5 HMNs) and LPA+ML-7 (n = 7 HMNs) treated groups of HMNs compared with their respective pretreatment controls (before). (E) eEPSCs AMPA evoked in HMNs by paired-pulse stimulation of VLRF before and following treatment with LPA and finally after coaddition of ML-7. (F) Changes in PPR of eEPSCs AMPA measured in HMNs exposed sequentially to LPA and LPA+ML-7. Next, we explored whether LPA modulates GABAergic and glutamatergic synapses by similar mechanisms of action. Amplitude, but not frequency, of miniature quantal IPSCs GABAA (mIPSCs GABAA ) recorded in HMNs was reduced by LPA, in agreement with a postsynaptic site of action ( Fig 6A; S9 Fig). The molecular cascade downstream of LPA is also distinct, since LPA-induced alterations on mIPSCs GABAA were reversed by the ROCK inhibitor H1152 ( Fig  6A; S9 Fig). H1152 also returned (s-)LPA-induced changes in eIPSC GABAA amplitude to a control-like state (S10A and S10B Fig). In support of a non-presynaptic action of s-LPA on eIPSCs GABAA , the mean PPR remained similar to the control condition in the presence of s-LPA or s-LPA plus H1152 (S10C and S10D Fig). Colocalization in HMNs of LPA 1 -ir with the postsynaptic marker gephyrin, a clustering protein for GABA A Rs [35], strengthened the evidence of a postsynaptic site of action for LPA ( Fig 6B).
Postsynaptic action and the molecular signaling underlying LPA-induced modulation of GABA A ergic system were assessed in primary cultures of spinal motoneurons (SMNs) (S1 Text; S11 Fig). The mean amplitude of inward GABA A R-mediated current evoked by exogenous GABA pulses (−4.13 ± 0.98 nA; n = 8 SMNs) was robustly reduced by s-LPA (−62.5 ± 10.1%, p < 0.001, one-way ANOVA for repeated measures (RM-ANOVA)), in a ROCK-dependent way (s-LPA+H1152: −3.23 ± 0.49 nA, p = 0.345) (Fig 6C). In addition, we observed that s-LPA activated the small GTP-binding protein RhoA, the major ROCK activator, in SMNs. This was evidenced by an s-LPA-induced increase (+78.3 ± 25.7%; p < 0.05, oneway ANOVA on Ranks) in the membrane (M):cytosolic (C) ratio of RhoA expression relative to the control condition ( Fig 6D). Supplementary data support LPA signaling as the activator for the RhoA/ROCK pathway in motoneurons (S1 Text; S12 Fig). Furthermore, pretreatment with siRNA lpa1 prevented the effects of (s-)LPA on GABA A R-mediated currents compared to cRNA-treated SMNs, providing conclusive evidence of postsynaptic LPA 1 involvement (Fig 6E  and 6F; S1 Text; S13 Fig).
Phosphorylation of serine 327 on the GABA Aγ2 subunit (pGABA Aγ2 ) regulates GABA A R clustering and synaptic strength at inhibitory synapses [36,37]. Therefore, we investigated whether LPA 1 -ROCK signaling regulates phosphorylation of GABA A γ 2 . Contrary to expectations of a direct interaction between ROCK and GABA Aγ2 , s-LPA induced a robust reduction (−83.3 ± 5.2%) of the pGABA Aγ2 :GABA Aγ2 ratio in SMNs that was prevented by coaddition of H1152 (+1.6 ± 6.0%) (Fig 6G). This was also observed in the HN (S1 Text; S14 Fig). Strikingly, direct binding of the phosphatase calcineurin (CaN) to GABA Aγ2 subunits dephosphorylates Ser 327 [37,38], which leads to a reduction in inhibitory postsynaptic current amplitude [37]. ; LPA plus ML-7, n = 102 boutons/a.z. *p < 0.05, one-way ANOVA relative to the control condition. Experiments and analysis were performed as in our previous published study [8]. Plots data can be found in S1 Data.  Therefore, recruitment of CaN (also named Ca 2+ /calmodulin-dependent phosphatase 2B), was proposed as a potential link between LPA 1 -ROCK signaling and GABA Aγ2 dephosphorylation.

LPA Induces Internalization of GABA Aγ2 Subunit
It is generally accepted that dephosphorylation appears to be important for receptor endocytosis [4,9]. As a next step, we investigated whether LPA-triggered dephosphorylation was accompanied by further subunit internalization. We found that s-LPA (15 min) led to a strong reduction (−99.9 ± 0.01%) in the amount of GABA Aγ2 allocated in M fraction in SMN cultures. A proportional increase (+109.4 ± 14.1%) in the quantity of GABA Aγ2 was observed in the C fraction relative to total GABA Aγ2 (Fig 7A). These outcomes suggest a translocation of at least this subunit from the SMN membrane to the cytosol triggered by s-LPA. The s-LPA-induced translocation was prevented by coincubation with either the ROCK inhibitor H1152 or the CaN inhibitor Cap (Fig 7A). GABA Aγ2 compartmentalization in SMNs was maintained after treatment with H1152 or Cap per se (Fig 7A).
To explore whether internalization is actually required for LPA-induced GABA A ergic STD, and given that GABA A R endocytosis is dynamin-dependent [39], we added the dynamin inhibitor dynasore to the bath to block GABA A R endocytosis. Dynasore (80 μM for 30 min) fully prevented both a reduction in the GABA Aγ2 M:T ratio and an increase in the C:T ratio induced by s-LPA, which was not altered by vehicle (−84.5 ± 5.8%). Dynasore per se did not modify GABA Aγ2 location (−6.4 ± 18.8%) relative to the vehicle condition (100.0 ± 36.7%) (Fig 7B). Interestingly, electrophysiological recordings showed that preincubation with dynasore had no effect on s-LPA-induced changes in GABA-evoked currents (−48.1 ± 8.7%; n = 4 SMNs) ( Fig  7C). These outcomes support that GABA Aγ2 internalization by endocytosis is not required for the attenuation in GABA A ergic neurotransmission induced by LPA signaling.
CaN-dependent dephosphorylation of Ser 327 at the GABA Aγ2 subunit is involved in the increase of lateral diffusion and cluster dispersal of surface GABA A Rs in the dendrites of cultured hippocampal neurons [36,40]. Therefore, we investigated whether s-LPA-induced STD under endocytosis inhibition conditions would involve GABA A R cluster disarrangement. Double immunolabelling for GABA Aγ2 and the postsynaptic scaffolding protein, gephyrin, confirmed GABA Aγ2 -ir clusters at the surface of SMNs, most of them colocalized with gephyrin-ir clusters (Fig 7D). In consonance with phospholipid-evoked GABA A R internalization, treatment with s-LPA (10 min) reduced mean fluorescence intensity, but not area, per cluster for these two postsynaptic proteins (Fig 7E-7G). However, the size of surface GABA Aγ2 -ir clusters increased in parallel with a reduction in fluorescence when s-LPA was added after pretreatment with dynasore (Fig 7E-7G). This agrees with s-LPA-induced lateral diffusion and cluster dispersal of GABA A Rs. In addition, the mean area of GABA Aγ2 -associated clusters of gephyrin was unaltered, but fluorescence was reduced by s-LPA under endocytosis inhibition (Fig 7E-7G). These Effects under the presence of dynasore support that this GABA A R disarrangement might involve previous lateral diffusion and cluster dispersal of surface GABA A Rs like that reported previously for cultured hippocampal neurons [36,40].
In summary, our data highlight a pathway by which, via recruitment of RhoA/ROCK signaling, postsynaptic LPA 1 evokes CaN-dependent dephosphorylation at Ser 327 of the GABA Aγ2 subunit, which is followed by GABA A R cluster dispersion and its concomitant translocation from the plasma membrane to the cytosol (Fig 7H). The latter does not seem to be required for the reduction in GABA A ergic synaptic strength triggered by LPA. Phospholipid-induced synaptic strength depression seems to be mainly supported by GABA Aγ2 dephosphorylation and subsequent GABA A R cluster dispersal.

LPA 1 Is Essential for Activity-Dependent Synaptic Plasticity
Next, the role of LPA signaling in short-term, activity-dependent synaptic plasticity was explored. N-methyl-D-aspartate receptor (NMDAR) activation causes a rapid, local, surface dispersal of synaptic GABA A Rs leading to an inhibitory synaptic depression [36,37]. We directly examined the role of LPA 1 -mediated signaling in NMDAR-induced STD of GABA A ergic signaling in SMNs. In cRNA-treated SMNs, perfusion of glutamate and glycine (Glut/Gly) for 4 min caused a rapid and reversible depression in GABA-induced current (−59.6 ± 5.3%, p < 0.001) in the presence of TTX, d-tubocurarine, strychnine and NBQX. This activity-dependent plastic event was absent in SMNs precultured with siRNA lpa1 (−15.2 ± 8.7%; Fig 8A), in untreated cells under zero extracellular Ca 2+ (−9.3 ± 11.5%; n = 6 SMNs), or in the presence of APV (−5.4 ± 13.9%; n = 6 SMNs), demonstrating Ca 2+ -and NMDAR-dependence. LPA 1 indicated treatments (n > 1,200 clusters per condition). *p < 0.005, Student's t test relative to control or dynasore condition. (H) Diagram of the proposed pathway mediating LPA-induced STD on GABA A Rmediated neurotransmission. Drug targets are also indicated. Plots data can be found in S1 Data.
doi:10.1371/journal.pbio.1002153.g007  Fig 6E, but GABA pulses were performed before and after 4 min addition to the bath of Glut (30 μM) and Gly (1 μM) and after successive washing (cRNA: n = 4 SMNs; siRNA lpa1 : n = 5 SMNs). *p < 0.05, one-way RM-ANOVA relative to the control (before) condition. (B) Same as in Fig 6G but performed from SMNs cultures receiving indicated pretreatments and incubated for 4 min with aCSF alone or with Glut/Gly to stimulate NMDARs. *p < 0.05, one-way ANOVA on Ranks relative to control (untreated) condition. Experiments were carried out in the presence of TTX, d-tubocurarine, strychnine, and NBQX. Plots data can be found in S1 Data.
doi:10.1371/journal.pbio.1002153.g008 knockdown reduced by approximately 40% the magnitude of activity-dependent STD at inhibitory synapses. From an extrapolation of these values to the dose-response curve in Fig 2C, it could be indirectly estimated that local concentrations of phospholipids achieved in response to those levels of motoneuron activity were first order micromolar, assuming all synthesized and released phospholipids were the monounsaturated form of LPA (18:1).
Glut/Gly also caused a drastic decrease in the pGABA Aγ2 :GABA Aγ2 ratio in untreated or cRNA-incubated SMNs, which was prevented by siRNA lpa1 (Fig 8B). Altogether, these data indicate that NMDAR-driven GABA-current depression was spike-independent and essential to extracellular Ca 2+ entry via NMDARs and LPA 1 activation, which downstream induces Ser 327 GABA Aγ2 dephosphorylation.
Findings from activity-dependent synaptic plasticity experiments agree with the notion that motoneurons are potential sources for Ca 2+ -dependent, spike-independent synthesis and release of lysophospholipids, which in turn might stimulate autocrine signaling pathways (to modulate inhibitory synapses), at least by way of the LPA 1 receptor. These outcomes also strongly point to lysophospholipids as paracrine retrograde messengers that act on presynaptic LPA 1 to regulate excitatory synapses; however, further research is needed to confirm this possibility.

Endogenous LPA Signaling Restrains Baseline Activity of Motoneurons in Adulthood
Finally, physiological involvement of LPA signaling in performance of motor output commands was investigated. In vivo, most HMNs exhibit rhythmic inspiratory-related bursting discharges driven by glutamatergic brain stem afferents, mainly acting on AMPARs, with little or no contribution of inhibitory inputs [22,23]. We began by analyzing the level and pattern of expression of the LPA 1 receptor within the HN of the adult rat. qRT-PCR analysis showed that disparity between lpa 1 and lpa 2-6 transcripts in the HN was even more accentuated in adults than at the neonatal stage (Fig 9A). Interestingly, mRNA and protein levels for LPA 1 at adulthood were approximately 150% and 140%, respectively, higher than in neonatal animals (Fig 9A and 9B). These results suggest a gain in relevance of LPA 1 -mediated signaling in the HN during postnatal development, supporting previous observations in the murine brain [41]. Immunohistochemistry revealed LPA 1 -ir puncta-like structures all along the HN (Fig 9C) and colocalization between VGLUT2and LPA 1 -ir puncta (Fig 9D, 9E, and 9H). A high proportion of VGLUT2-ir inputs (47.9 ± 3.4%; n = 55 HMNs) apposed to the perikarya of SMI32-identified HMNs were colocalizing with LPA 1 -ir puncta (Fig 9E and 9F). This also supposed an increase of approximately 150% during postnatal maturation. LPA 1 -ir appeared to border and colocalize with SMI32-ir structures (Fig 9G), supporting cytoplasmic and membrane location of LPA 1 in adult HMNs. Therefore, the molecular machinery to support a role of LPA 1 in modulating excitatory neurotransmission is also present in adults.
Additionally, in vivo decerebrated rats maintain respiratory activity [22,23]. To look for a role of LPA signaling in processing motoneuron inspiratory activity, LPA 1/3 inhibitors VPC 32179 (0.5 mM), VPC 32183 (1 mM), and Ki16425 (2 mM) or its vehicle (10% DMSO) were microiontophoretically applied to antidromically-identified HMNs subjected to unitary extracellular recordings (Fig 9I). The effect of these drugs on the unitary basal firing inspiratory-related activity of HMNs in basal conditions (end-tidal CO 2 = 4.8%-5.2%) was evaluated. The time course of the mean firing rate averaged over the duration of the inspiratory burst (mFR/ burst) was measured by applying increasing currents (−20 to −140 nA, 30 s duration) through the drug barrels (Fig 9J-9L). A current-dependent increase in the mFR/burst of HMNs was Unitary discharge activity (Rec) of HMNs was obtained in decerebrated, vagotomized, and artificially ventilated adult rats, which had been injected with a neuromuscular blocking agent. A three-barreled pipette with a barrel for electrophysiological recordings and another for microiontophoretic administration of a drug are illustrated. HMNs were identified by their antidromic activation from the electrode (St.) implanted in the XIIth nerve and by the collision test (top traces) between spontaneous orthodromic (dot) and antidromic (asterisk) evoked action potentials. When the stimulus was triggered by a spontaneous spike at a short latency, the antidromic action potential was occluded (arrowhead). Middle and bottom traces represent the extracellularly recorded spike discharge for an inspiratory HMN and the histogram of the instantaneous firing rate (FR, in spikes (sp)/s), respectively. Mean observed for all drugs but not when current was applied to the vehicle solution (Fig 9J-9L). In summary, these data point to a physiological role for LPA signaling in motor output performance by restraining the inspiratory-related activity driven by glutamatergic inputs to HMNs.

Discussion
The present study showed that bioactive membrane-derived phospholipids evoke rapid and reversible synaptic depression and mediate activity-dependent synaptic plasticity, mainly via LPA 1 . Phospholipids likely operate as local messengers in activity-dependent GABAergic STD in a Ca 2+ -dependent, spike-independent manner. Strikingly, at physiological concentrations of nanomolar to first order micromolar, LPA has a greater effect on inhibitory than excitatory inputs. Finally, LPA signaling regulates brain-elemental processing tasks such as performance of motor output commands. These data open a new scenario in which the membrane-phospholipid metabolism actively participates in controlling synaptic strength, and then affects neuronal excitability in physiological and pathological states.
Important determinants of synaptic strength, short-term plasticity and intersynaptic crosstalk mainly involve fine-tuning of the number of neurotransmitter receptors and the RRP size of SVs [4,8]. LPA depresses the main excitatory and inhibitory synaptic systems, affecting both by different degrees, loci, and mechanisms of action. At glutamatergic synapses, and by way of presynaptic G αi/o -protein-coupled LPA 1 and PLC-MLCK activation, LPA results in MLC phosphorylation, which might stimulate the actomyosin contractile apparatus [32] to reduce the bulk of the RRP of SVs (Fig 10). Depletion of some RRP of SVs usually underlies short-term forms of synaptic depression [1,2]. Ultrastructural correlates for LPA-induced STD further supported that functional synaptic changes are partly explained by a reduction in the size of the RRP of SVs. Changes in the actin cytoskeleton are a prerequisite for exocytosis, enabling docking and fusion of SVs with the plasmalemma [32]. As in our results, LPA-dependent contraction of smooth muscle cells involves activation of PLC and MLCK, followed by MLC phosphorylation [31] that promotes actomyosin interactions [32]. In this context, a physical relationship between p-MLC and glutamatergic synapses on adult and neonatal motoneurons has been recently reported [42]. At the calyx of Held synapse, MLCK controls the size of the fast-releasing pool of SVs [43]. In addition, ROCK regulates p-MLC levels via MLCK inhibition to maintain basal RRP ordering of SVs at excitatory inputs [8,42]. Therefore, presynaptic LPAdependent and ROCK signaling seem to converge onto a common molecular mechanism, namely MLC phosphorylation and size of the RRP at excitatory synapses. It is interesting, then, that the ROCK inhibitor did not actually enhance LPA-induced depression of AMPAR currents. These outcomes suggest that the antagonistic functional actions of ROCK and LPA 1 -signaling, converging on MLCK, results in a push-pull mechanism that regulates the size of the RRP of SVs at excitatory synapses.
At GABAergic synapses, LPA dephosphorylates Ser 327 of GABA Aγ2 subunits and favors GABA Aγ2 internalization via postsynaptic G α12/13 -coupled LPA 1 /RhoA/ROCK signaling and firing rate (mFR, red dotted line) in each burst was measured and subsequently plotted along time. (J) Instantaneous firing rates (sp/s) of two HMNs in response to microiontophoretic administration of VPC 32183 or vehicle (10% DMSO in PBS, pH 8.0) at the indicated current. During the before condition, a retention current of +5 nA is continuously applied. Note the lack of effect of vehicle and the stimulating effect exerted by the application of the LPA 1/3 inhibitor. (K) Time course of the mean FR (mFR, sp/s) per burst in response to microiontophoretic administration (30 s on, 60 s off) of VPC 32183 or vehicle at the indicated applied currents.
doi:10.1371/journal.pbio.1002153.g009 subsequent CaN activation (Fig 10). The cell surface stability of GABA A Rs is regulated by posttranslational modifications such as phosphorylation. GABA A R phosphorylation is involved in the modulation of receptor biophysical properties and membrane trafficking [44]. Phosphorylation stabilizes the GABA A R on the surface and, conversely, dephosphorylation is important for receptor endocytosis [4]. NMDAR activation causes GABA A R cluster dispersal and lateral diffusion by CaN activation and dephosphorylation of Ser 327 GABA A γ 2 [36,40], leading to long-term depression at CA1 inhibitory synapses [37]. Dispersal could involve receptor clustering at clathrin-coated sites at the plasmalemma, which invaginate and pinch off to form clathrin-coated vesicles. Internalized receptors are then either subject to rapid recycling or are targeted for lysosomal degradation [4].
Our results indicated that the LPA 1 -RhoA/ROCK-CaN pathway dephosphorylates the GABA A γ 2 subunit, which undergoes lateral diffusion, dispersal of clusters, and subsequent endocytosis (Fig 10). However, endocytosis does not seem to be crucial for LPA-induced functional depression at GABA A ergic neurotransmission, which seemed to be mainly supported by GABA A γ 2 dephosphorylation and subsequent clusters dispersal of surface GABA A Rs. The kinetic recovery suggests rapid replenishment of the synaptic GABA A R content, given that re-establishment of inhibitory synaptic strength occurred with 7 to 10 min washing after LPAinduced depression. The coordinated action of kinases and phosphatases, downstream of LPA 1 -triggered signaling, then plays a pivotal role in controlling neuronal excitability by modulation of GABA A γ 2 phosphorylation and receptor recycling.
The present results seem controversial in relation to our previous findings demonstrating a presynaptic role for endogenous baseline ROCK activity in the regulation of AMPAergic and GABA A ergic neurotransmission [8]; here, we describe that ROCK also acts postsynaptically to mediate LPA-induced depression of the GABA A ergic transmission. Whether presynaptic baseline ROCK activity in inhibitory inputs depends on membrane-derived bioactive lipid mediators, such as LPA and/or sphingosine 1-phosphate, remains to be elucidated. Nevertheless, at glutamatergic synapses, ROCK activity is likely independent of LPA 1/3 signaling, because inhibitors of these receptors did not mimic AMPAergic STD induced by ROCK inhibition. However, we cannot discard the involvement of another LPAR in maintaining baseline ROCK activity in the synaptic terminals. Interestingly, although presynaptic ROCK is active in our experimental conditions [8], postsynaptic endogenous activity of ROCK, if any, is even below the level required to reveal its impact on synaptic strength and membrane properties [8] of motoneurons. This could be explained by the differential expression of ROCK isoforms at the two compartments, ROCKα in the postsynaptic site and ROCKβ in the presynaptic one, and/or the lower concentration of ROCKα in motoneurons relative to synaptic structures [8]. Anyway, data suggest that when motoneuron activity is low, presynaptic ROCK activity maintains inhibitory synaptic strength by stabilizing the size of the RRP of SVs. However, after exogenous addition of LPA or when motoneuron activity rises, and subsequent coupled LPA synthesis and/or release occurs, postsynaptic LPA 1 stimulates ROCK. This leads to deinhibition by GABA Aγ2 dephosphorylation and receptor endocytosis.
In the rat, the highest LPA concentration in tissue is found in the brain [12]. Cultured cortical neurons produce LPA at nanomolar concentrations [45], but LPA levels increase up to 10 μM after injury, trauma, or hemorrhage involving blood-brain barrier damage [46]. Here, physiological concentrations (nanomolar to first order micromolar) of LPA affected GABAergic to a greater degree than glutamatergic inputs, achieving maximal and similar affectation at 10 μM. Thus, it is possible that LPA signaling maintains neuronal excitability around a dynamic range, promoting deinhibition at low levels of neuronal activity and depressing excitatory inputs when activity increases, perhaps as part of a homeostatic mechanism that prevents excitotoxicity. Any candidate for coupling synaptic strength to neuronal activity must be regulated by activity at the postsynaptic site. Interestingly, noxious stimulation of primary afferent neurons induces LPA production in the dorsal horn in a glutamate-dependent manner [21]. Here, LPA signaling, mainly via LPA 1 , was essential in STD of inhibitory inputs triggered by precedent activity of the neuron. Autocrine LPA signaling was essential for NMDAR-driven GABA-current depression, which depends on extracellular Ca 2+ entry passing through NMDARs. Activity-dependent synaptic plasticity occurred independently of the generation of action potentials at the postsynaptic neuron. Postsynaptic [Ca 2+ ] increase and LPA signaling dependence for activity-dependent STD in cultured motoneurons strongly support that this cell type is a potential source for activity-dependent LPA synthesis and/or release.
Despite the apparent lack of endogenous LPA signaling affecting synaptic strength in our in vitro model, local iontophoretic application of three LPA 1/3 inhibitors increased, in a dose-dependent manner, the baseline inspiratory-related activity of HMNs in the adult rat. This rhythmic inspiratory-related bursting discharge of HMNs is driven mainly by glutamatergic brain stem afferences, with little or no contribution of inhibitory inputs [22,47]. There is an apparent gain in relevance of LPA 1 -mediated signaling in the HN during postnatal development, to the detriment of LPA 2-6 -triggered pathways, as well as excitatory inputs apposed to adult HMNs express LPA 1 . Taken together, these findings support that phospholipids, most likely activating LPA 1 at glutamatergic synapses, controlled physiological inspiratory-related activity of HMNs, presumably by restraining their AMPAergic input drive [22]. Thus, endogenous LPA signaling physiologically contributes in the performance of normal patterns of motor output commands in adult animals.
Alterations in phospholipid homeostasis affect various pathological conditions, thus attracting increased diagnostic and pharmacological interest [48]. The exquisite balance between excitatory and inhibitory inputs is critical for the proper functioning of the brain, and its imbalance leads to the cognitive impairment associated with neurodegenerative diseases and metabolic syndromes related to obesity, dyslipidemia, lipodystrophy, insulin resistance, and alcoholism [49][50][51]. In particular, LPA production and/or autotaxin are increased in obesity-associated metabolic diseases [52], induced hypercholesterolemia [53], congenital lipodystrophy [54], as well as in ethanol-fed mice [55] and in patients with Alzheimer disease [56] or multiple sclerosis [57]. In addition, phospholipids uptake in mammalian cells depends on their activation status, a critical support for cellular incorporation of nutrition-derived fatty acids. Imported phospholipids are utilized for production of bioactive lipids, such as LPA [58], and thereby modify synaptic transmission. Therefore, we can point to LPA as a promising candidate in coupling brain function, by modulating synaptic strength and plasticity, to the metabolic condition of the organism across physiological and pathological states.

Materials and Methods
Wistar rats of either sex and CD1 pregnant mice were obtained from an authorized supplier (Animal Supply Services, University of Cádiz, Spain), and were cared for and handled in accordance with the guidelines of the European Union Council (86/609/UE) and Spanish regulations (BOE 67/8509-12; BOE 1201/2005) on the use of laboratory animals. Animals were individually housed-except neonatal animals, which were housed with their mother-in cages with water and food pellets available ad libitum, under temperature-controlled conditions at 21 ± 1°C, with a 12 h light and dark cycle. Efforts were made to minimize the number of animals used and their suffering. All surgical procedures were carried out under aseptic conditions. Experimental procedures were approved by the local Animal Care and Ethics Committee.

Electrophysiological Recordings
In vitro whole-cell patch-clamp recordings of motoneurons. Whole-cell patch-clamp experiments were performed on cultured SMNs or on HMNs from transverse brain stem slices (300-400 μm thick) of P6-P9 rats as previously described [8,42,59]. Whole-cell AMPAergic responses were recorded at a holding potential of −65 mV with the KGluconate-based intracellular solution. GABA A postsynaptic currents were recorded in cells voltage-clamped at −75 mV using the CsCl-based electrode solution. The AMPAergic or GABA A ergic component of the evoked currents was pharmacologically isolated as indicated in the legend of Fig 2B. Unitary extracellular recordings of HMNs in the adult rat. Adult animals (250-300 gr) were prepared for extracellular recordings as reported previously [60,61]. Tracheotomized, vagotomized, and decerebrated animals were paralyzed and mechanically ventilated. End-tidal CO 2 was kept at 4.8%-5.2% along the recording session. Three-barreled, microfilament-filled glass pipettes were used for single-unit recording and microiontophoretic drug administration.

Electron Microscopy
Brain stem slices (300 μm thick) incubated for 10 min (approximately 22°C), with aCSF alone, 0.2% DMSO (vehicle) or with various drug treatments were immediately fixed and processed for electron microscopy analysis. Ultrathin sections (70-80 nm thick) were analyzed at high magnification (43,000x). Only boutons, contacting with motoneurons at the level of the nucleolus, evidencing at least an a.z. were included in this study [8].

siRNA-Mediated Silencing of lpa1
Neonatal rats (P4) received an acute injection of siRNA lpa1, or nontargeting siRNA (cRNA), (2 μg/rat) in 2 μl of RNase-free PBS into the fourth ventricle. The target sequence for the siR-NA lpa1 was UCAUUGUGCUUGGUGCCUU. A group of animals was infused with 2 μl of RNase-free PBS (vehicle) as an additional control. Primary cultures of SMNs were incubated with 2.5 μl of either cRNA or siRNA lpa1 (each 100 μM) for 72 h at 37°C. Cells were then collected for qRT-PCR analyses or used for electrophysiological studies.

Quantitative Real-Time Reverse Transcriptase PCR (qRT-PCR)
Total RNA was extracted from the HN or cultured SMNs using TRIzol, and 0.5 μg of RNA was used for cDNA synthesis with iScript cDNA synthesis. The PCR primers were as indicated in S2 Table. Western Blotting Total protein was extracted from microdissected HNs, NSC34 cells, and membrane and cytosol fractions of NSC34 cells and SMNs. Membranes were blotted with specific antibodies against GABA Aγ2 , p Ser327 GABA Aγ2 , LPA 1 , p-MLC, MLC, or RhoA. Membranes were also probed with anti-α 1 -tubulin or anti-β-actin antibodies as control for the total amount of protein contained in each well.

Statistics and Data Analysis
Data are expressed as the mean ± standard error of the mean (SEM). The number of analyzed specimens per experimental condition is indicated in figure legends or in the result section. Data were obtained from at least three animals per experimental condition. In ROCK activity, western blotting and qRT-PCR experiments, each individual assay was performed by using tissue samples collected from at least six animals per experimental condition. Quantitative data from ROCK and CaN activity assays, western blot, and qRT-PCR represent the average of, at least, three independent experiments. Applied statistical tests per experimental condition are indicated in figure legends or in results. Post hoc Holm Sidak or Dunn tests were applied for ANOVA for repeated measures or on Ranks, respectively. In all cases, the minimum significance level was set at p < 0.05. HMNs recorded under indicated conditions. Bin width: 2 pA. Plot data can be found in S1 Data. (C) Whole-cell AMPAergic currents evoked by 100 ms pressure pulses of glutamate (applied at saturating concentrations; 1 mM) in a HMN before and after superfusion with LPA. Recordings were performed in the presence of TTX in nominally Ca 2+ -free solution. Experiments and analysis were performed as in our previously published study [8]. Top, examples of eEPSCs AMPA recorded in a HMN in response to paired-pulse stimulation of VLRF axons at the indicated conditions. Stimulus interval was 25 ms. The rightmost trace shows the superimposition of the responses scaled to the peak of the first eEPSCs AMPA . Bottom, PPR was obtained from the amplitude of the first and second eEPSCs AMPA by the formula eEPSCs AMPA 2/eEPSCs AMPA 1. Comparison of PPR measured at interpulse intervals ranging from 25 to 200 ms for HMNs recorded before, during, and after washout of the s-LPA (40 μM; n = 6 HMNs). Ã p < 0.05, two-way RM-ANOVA relative to control (before) condition. Experiments and analysis were performed as in our previously published study [8]. Plots data can be found in S1 Data. HMNs were initially allowed to stabilize (Stabil.) with normal aCSF to obtain baseline control recordings. Slices were then superfused for 10 min with aCSF supplemented with 0.2% DMSO, the LPA 1/3 inhibitor Ki16425 (0.4 μM in 0.2% DMSO; Drug, left protocol) or LPA (2.5 μM; Drug-1, right protocol) before current responses were acquired again. In the right protocol, slices were additionally incubated for 10 min with LPA plus DMSO or with Ki16425 (0.4 μM; Drug-2). Finally, a last round of acquisition was taken after a 10 min washout with drug-free aCSF. Bottom, representative eEPSCs AMPA from HMNs recorded at the indicated conditions. (B, C) Mean eEPSCs AMPA amplitude (B) and PPR ratio (C) measured at 25 ms interpulse intervals for HMNs recorded under the indicated treatments (n ! 5 HMNs per condition). Ã p < 0.05, one-way RM-ANOVA relative to control (before) condition. Plots data can be found in S1 Data. (TIF) S7 Fig. Effectiveness of siRNA lpa1 in knockdown LPA 1 in the brain stem of neonatal rats. (A) Expression levels of mRNA for indicated LPARs obtained by qRT-PCR of isolated brain stems at P6 after receiving the indicated treatments at P4. GAPDH was used as housekeeping. Values were normalized taking control condition (untreated animals) as 1. Ã p < 0.05, one-way ANOVA on Ranks relative to control, vehicle, and cRNA conditions for each receptor. (B, C) Immunohistochemistry against LPA 1 of brain stem coronal hemisections obtained from P6 pups untreated (Control), or receiving the indicated treatments at P4. Scale bar: 500 μm. Plot data can be found in S1 Data.  (1 μM; B), or the G αq/11 inhibitor YM-254890 (1 μM; C). Stimulus interval was 25 ms. (D, E) Mean eEPSCs AMPA amplitude reduction (D) and PPR ratio increase (E) measured at 25 ms interpulse intervals for HMNs recorded under the indicated treatments (n ! 4 HMNs per condition). Ã p < 0.05, one-way ANOVA relative to control condition. Plots data can be found in S1 Data. Amplitude distribution histograms (A) and cumulative probability functions (B) of mIPSCs GA-BAA at the indicated conditions. Each condition is represented by 600 events (5 pA bin width) pooled from 5 HMNs. Note that H1152 reversed the LPA-induced shift to the left of the distribution histograms and the cumulative probability functions of mIPSCs GABAA amplitude (p < 0.05; Kolmogorov-Smirnov test). Plots data can be found in S1 Data. (TIF) S10 Fig. Evidence for a non-presynaptic mechanism underlying LPA-ROCK-induced depression of GABA A ergic neurotransmission. (A, B) Illustrative eIPSCs GABAA of two HMNs (A) and summary data of eIPSCs GABAA amplitude (B) recorded before and after LPA (2.5 μM; left) or s-LPA (40 μM; right) treatment, after the next coaddition of H1152 (20 μM) and subsequent washing (n = 4 HMNs). Ã p < 0.05, one-way RM-ANOVA relative to the control (before) condition. (C, D) Examples of eIPSCs GABAA recorded in a HMN (C) in response to pairedpulse stimulation of VLRF axons and changes in PPR (D) (n = 4 HMNs). Plots data can be found in S1 Data. (TIF) S11 Fig. SMNs express LPA 1 . (A) Epifluorescence images of cultured SMNs processed by immunohistochemistry for the motoneuron marker SMI32 (left) and counterstained with the nuclear marker DAPI (right). Note that all cells in the field are SMI32-ir. (B) Expression levels of mRNA for the indicated LPARs obtained by qRT-PCR of cultured SMNs relative to the housekeeping GAPDH. Ã p < 0.05, one-way ANOVA on Ranks relative to lpa 2-6 . (C) Epifluorescence images of cultured SMNs processed by immunohistochemistry for SMI32 (top) and LPA 1 (bottom). Scale bars: A, 25 μm; C, 100 μm. Plot data can be found in S1 Data. (TIF) S12 Fig. (s-)LPA stimulates RhoA/ROCK signaling in motoneurons. (A) Left, western blots of LPA 1 and total (T), cytosolic (C), and membrane-associated (M) RhoA in the motoneuronlike cell line NSC34 after indicated treatments. For LPA 1 , the cell line HEK293 was taken as a negative control and β-actin expression was used as an internal loading reference. Right, histogram showing the average ratio of densitometric intensity in M or C fractions relative to total RhoA at the indicated conditions. Ratio values were normalized relative to the control group. (B, C) Summary histogram of changes in ROCK activity in homogenates from HN (B) and cultured NSC34 (C) untreated (control) or treated with either s-LPA (40 μM), H1152 (100 μM), or s-LPA plus H1152. Ã , # p < 0.05, one-way ANOVA on Ranks relative to the control and both control and s-LPA-treated groups, respectively. Plots data can be found in S1 Data. (TIF) S13 Fig. Effectiveness of siRNA lpa1 in knockdown LPA 1 in SMNs. (A) Expression levels of LPAR mRNAs in SMNs after incubation with the small interfering RNA against lpa 1 (siR-NA lpa1 ) relative to cultures treated with a nontargeting siRNA (cRNA). Ã p < 0.05, one-way ANOVA on Ranks relative to lpa 2-6 . (B, C) Epifluorescence images of cultured SMNs receiving the indicated treatments processed by immunohistochemistry for SMI32 and LPA 1 . Immunohistochemical processing was performed in parallel. Scale bars: 25 μm. Plot data can be found in S1 Data. (TIF) S14 Fig. s-LPA induces GABA A γ 2 dephosphorylation in the HN by a ROCK-dependent mechanism. Western blot (top) and averaged ratio (bottom) of phosphorylated and total GABA A γ 2 subunit protein levels (denoted as pGABA A γ 2 and GABA A γ 2 , respectively) in the HN of neonatal brain stem slices incubated (10 min) with aCSF alone (control) or supplemented with indicated drugs. β-actin was an internal loading reference. Ã p < 0.05, one-way ANOVA on Ranks relative to control condition. Plot data can be found in S1 Data.