Animals

Twelve-week-old C57BL/6 male mice (n = 4) were used for each qualitative experiment and n = 6 mice for each quantification experiment described in this study. An orexin/enhanced green fluorescent protein (Ox-EGFP) breeder pair was a kind gift from Prof. Takeshi Sakurai (Kanazawa University, Japan). Brain sections from these mice show green fluorescence in the Ox neurons when excited at 488 nm wavelength. Breeding was set up in-house and the male pups were aged to ten-twelve weeks before the start of the experiments. Animals were maintained on a 12/12 h light/dark cycle (lights on at 8∶00 A.M., off at 8∶00 P.M.), in a temperature-controlled room (21.5±1°C). Animals received food and water ad libitum. All animal experiments were licenced by the UK Home Office in accordance with the Animals (Scientific Procedures) Act of 1986 and in agreement with UK and EU laws.

Perfusion, Fixation and Sectioning

Mice were anaesthetised with pentobarbitone (0.3 ml; Euthanal, Bayer AG, Leverkusen, Germany) and perfused transcardially with 0.1 M PBS (pH 7.4) buffer followed by ice-cold 4% paraformaldehyde in PBS. The brains were removed and fixed in 4% paraformaldehyde for at least 24 hr and cryoprotected in a 30% sucrose solution in PBS overnight at 4°C. Brains were then snap frozen using Envirofreez, and coronal sections of 45 µm thickness were cut using a Leica cryostat. According to the mouse brain atlas by Paxinos and Franklin (2004), LHA and DRN sections were cut at a depth of −0.34 mm to −2.80 mm bregma and −4.04 mm to −5.20 mm bregma respectively. Sections were chosen according to the stereological rules, with the first section taken at random and every sixth section afterward. In the case of LHA, by taking every 6^th section (in total 54 sections per LHA per brain half), at least 10–11 sections were taken per immunostaining experiment (n = 4). In the case of DRN, by taking every 3^rd section (in total 25 sections per DRN per full brain), at least 11–12 sections were taken per immunostaining experiment (n = 4). 10–11 sections from 4 mice brain halves (in case of LHA) and full brain (in case of DRN) were processed so, at least 40 sections were considered in total for every immunostaining experiment.

Immunohistochemistry

Single, double or triple immunofluorescence (IF) staining experiments were performed on 45 µm free-floating sections using primary antibodies: i) affinity purified goat polyclonal Ox-A (C-19) IgG (1∶400 dilution, Santa Cruz, sc-8070), raised against a peptide mapping at the C-terminus of Orexin-A of human origin; ii) rabbit polyclonal anti-5HT_1B receptor IgG (1∶500 dilution, Abcam, ab102700) raised against a synthetic peptide taken from within the region 230–280 (designed to the 3rd cytoplasmic domain) of the human 5HT_1B receptor conjugated to an immunogenic carrier protein; iii) rabbit polyclonal anti-5HT_1A receptor IgG (1∶500 dilution, Abcam, ab79230) raised against a synthetic peptide from the 3rd cytoplasmic domain of mouse 5HT_1A receptor, conjugated to an immunogenic carrier protein [44]; iv) rabbit polyclonal anti-5HT_2A receptor IgG (1∶500 dilution, Abcam, ab16028) raised against a synthetic peptide conjugated to KLH derived from within residues 1–100 of rat 5HT_2A receptor [45], [46], [47], [48], [49]; v) rabbit polyclonal anti-5HT_2C receptor IgG (1∶500 dilution, Abcam, ab32172) raised against a synthetic peptide derived from within residues 400 to the C-terminus of rat 5HT2C receptor [48], [50], [51], [52]; vi) rabbit polyclonal anti-HTR3A receptor IgG (1∶500 dilution, Sigma-Aldrich, AV13046) raised against a synthetic peptide corresponding to a region of human HTR3A with an internal ID of P01375. The immunogen for anti-HTR3A antibody was a synthetic peptide directed towards the N-terminal of human HTR3A, with the following sequence: LLWVQQALLALLLPTLLAQGEARRSRNTTRPALLRLSDYLLTNYRKVGRP; vii) rabbit monoclonal anti-NMDAR, clone 1.17.2.6 IgG (1∶500 dilution, Millipore, AB9864R) raised against linear peptide corresponding to human NMDAR1; viii) guinea pig polyclonal anti-GABA IgG (1∶400 dilution, Millipore, AB175) raised against GABA coupled to KLH via glutaraldehyde; ix) rabbit polyclonal anti-Orexin receptor 1 IgG (1∶500 dilution, Abcam, ab83960) raised against a synthetic peptide from the internal region (250–300) of mouse Orexin receptor 1 conjugated to an immunogenic carrier; x) rabbit polyclonal anti-Orexin receptor 2 IgG (1∶500 dilution, Abcam, ab85899) raised against a synthetic peptide from the C terminal region (350–415) of human Orexin receptor 2 conjugated to an immunogenic carrier; xi) rabbit monoclonal PSD-95 (D27E11) XP IgG (1∶500 dilution, Cell Signaling Technology, 3450S) raised against a synthetic peptide corresponding to residues surrounding Gly99 of human PSD95; xii) rabbit polyclonal Anti-Gephyrin IgG (1∶500 dilution, Abcam, ab32206) raised against a synthetic peptide conjugated to KLH derived from within residues 700 to the C-terminus of mouse Gephyrin.

All primary antibodies used in this study were previously characterized and their specificity was verified according to the respective manufacturer. After blocking in 1% BSA and 5% donkey normal serum in TBS buffer (pH 8, Sigma Aldrich) to avoid nonspecific antibody binding, sections were incubated in the primary antibody overnight at 4°C. The following day, sections were incubated in the secondary antibody for an hour at room temperature and mounted using Vectashield mounting medium (Vector Laboratories) on the slides coated with 3-aminopropyl tri-ethoxy silane (Sigma Aldrich). For double IF stainings, a simultaneous method was used where sections were incubated with two primary antibodies together for 48 hrs at 4°C. In the case of triple IF stainings, three primary antibodies were added together and sections were incubated for 72 hrs at 4°C. Labeled donkey IgG (H+L) anti-goat Alexa Fluor 488 (1∶800 dilution, Cat. # A11055, Molecular Probes), anti-rabbit Alexa Fluor 546 (1∶1000 dilution, Cat. # A11056, Molecular Probes), anti-chicken CF 594 (1∶1000 dilution, Cat. # BTIU20167, Biotium) and anti-guinea pig CF 633 (1∶1000 dilution, Cat. # BTIU20171, Biotium) secondary antibodies were used in the study. Negative controls were performed for single IF staining by omitting the primary antibody and for double/triple IF staining by omitting the primary and secondary antibody.

Antigen Retrieval

Antigen retrieval was done while performing NMDAR1 IF staining. Sections were incubated in 10 mM sodium citrate (pH 6) at 80°C for 30 min, before blocking the sections.

Microscopy

Co-localization Quantification

For quantification of the 5HT receptor subtype on the Ox neurons in LHA, images were acquired using 63× objective in a Leica SP5 confocal microscope. Once the conditions such as photomultiplier gain for each channel and pinhole settings were adjusted to minimize background noise and saturated pixels, parameters were kept constant for all acquisitions. Triple-stained images were obtained by sequential scanning for each channel to eliminate the cross talk of chromophores and to ensure reliable quantification of co-localization. Ambiguity and inconsistency are the two major issues affecting colocalization analysis. In the context of digital imaging, colocalization means the colours emitted by the fluorescent molecules occupy the same pixel in the image [53], [54]. Therefore, we have used the JACoP (Just Another Colocalization Plug-in) tool of Image J for colocalization analysis. The degree of Ox neuron (Alexa488, green) signal colocalizing with 5HT receptor (Alexa546, red) signal was quantified on single-plane 8-bit color images using the JACoP plugin [55]. A simple way of measuring the dependency of pixels in dual-channel images is to plot the pixel grey values of two images against each other. The intensity of a given pixel in the green image is used as the x-coordinate of the scatter plot and the intensity of the corresponding pixel in the red image as the y-coordinate. Results in JACoP are displayed in a pixel distribution diagram called a scatter plot or fluorogram in addition to the calculated co-localization coefficients such as Pearson’s and Overlap coefficient.

Pearson’s correlation coefficient (Rr) is the most quantitative estimate of colocalization that depends on the amount of colocalized signals in both channels in a nonlinear manner and is a well-defined and commonly accepted means for describing the extent of overlap between image pairs [56]. It is used for describing the correlation of the intensity distributions between channels. It takes into consideration only similarity between shapes, while ignoring the intensities of signals. The values of Pearson’s coefficient range from −1 to 1, with values from 0.5 to 1.0 indicating colocalization and −1.0 to 0.5 indicating no-colocalization. As Pearson’s Correlation does some averaging of pixel information and can return negative values another method, the Overlap Coefficient, is simultaneously used to describe overlap. Manders’ overlap coefficient (R) is based on the Pearson’s correlation coefficient with average intensity values being taken out of the mathematical expression. This new coefficient varies from 0 to 1, with values from 0.6 to 1.0 indicating colocalization and 0 to 0.6 indicating no colocalization. Overlap coefficient according to Manders indicates an overlap of the signals and thus represents the true degree of Colocalization. Costes randomization (number of randomization rounds = 1000) was used to exclude any co-localization of pixels that might have occurred due to chance [57]. P-value for each image pair was 100.0% (calculated from the fitted data).

Statistics

Statistical analyses were performed using Prism 5 (GraphPad software Inc. USA) with the level of probability set at 95% and the results are expressed as means±SEM. Data for 5-HT receptor quantification was analysed by two-tailed unpaired t-test.

Computational Model

To investigate the consequences of the DRN-LHA circuit architecture on systems dynamics, we implement neural network model that is an extension and modification of our previous model [58]. The aim of the model is to understand how the circuit connectivity and timescale affect the DRN-LHA activity.

Neural units.

Our neural network model consists of 4 populations of neurons, namely, Ox neurons in the LHA, local LHA inhibitory GABAergic neurons, 5-HT neurons in the DRN, and local DRN inhibitory GABAergic neurons. Glutamatergic neurons will be ignored in this work primarily due to the evidence showing that glutamatergic effects in DRN are locally weaker when compared to local GABAergic influence [59], and that we can implicitly encompass the effects of the LHA’s glutamatergic neurons on Ox neurons [4] with self-excitatory Ox connections. Furthermore, incorporating two additional (glutamatergic) neural populations can lead to more free parameters in the model.

The chosen model is of the population-averaged or “mean-field” firing rate type model [60], [61], [62], [63]. This simplifies a population of neurons into its representative unit. The 4 neural populations to be considered are the LHA’s Ox and GABAergic neural populations, and the DRN’s 5-HT and GABAergic neural populations. With support from electrophysiological data, the input-output or current-frequency relationship (f-I curve) for each neural population can be described by threshold-linear functions [64], [65]:

(1)

where f_j and (I_Local,j – I_Background,j) denote the population-averaged firing rate activity and total afferent input of the j^th neural population, respectively. I_Background,j is the background current, consisting of inputs from the rest of the brain areas. g_j and I_0,j determine the input-output slope and the current threshold, respectively. The threshold-linear function [z]₊ = z if z>0, and 0 otherwise. Based on their known neuronal electrophysiological properties [4], [27], [66]), we determine and constrain the values of the g_j’s (i.e. g_5-HT and g_GABA for DRN; g_GABA and g_Ox for LHA) and the I_0,j’s, while consider I_Background,j as a free parameter. Note that we have assumed the dynamics of the neural population (neuronal membrane time constant ∼ 10 ms) to be relatively instantaneous and slaved to the much slower timescale of the connections (∼ s) [63]. This simplification reduces 4 model parameters (neuronal membrane time constants) and 4 dynamical (differential) equations for the 4 neuronal types [60].

2.8.2. Inputs and connections: Since each neural unit receives inputs from self-feedback and from 2 other neural units (e.g. 5-HT neuronal population receives both afferent inputs from DRN’s GABAergic neurons and longer range connection from Ox neurons), the local afferent input (minus the background input) can be described by:(2)

where the subscript “self” denotes a self-feedback connection (e.g. due to autoreceptors in Ox and 5-HT neural populations, or GABAergic synapses in the two inhibitory neural populations), and the subscript “1″ or “2″ denotes the afferent inputs due to the other 2 neural populations. For local GABAergic neurons, they receive self-inhibition, and projections from their local principal neural population (5-HT or Ox if from DRN or LHA, respectively), and long-range projection from the other brain region (LHA or DRN). If the net effect of the i^th neural population on the j^th neural population is inhibitory or excitatory, the coefficient in front of I_j,i will be −1 or +1, respectively. For example, suppose the 5-HT neural population receives inhibitory autoreceptor influence, direct projection from Ox neurons, and local GABAergic influence. Then,(3)

To simulate the phasic response of the circuit, an extra term I_stim is added to the 5-HT or/and Ox inputs (4)

The dynamics of each input are filtered by its (synaptic/effective) time constant (τ_{syn, j, i}) as follows: (5)

In principle, there are 10 such similar dynamical equations describing the currents caused by ionotropic (4 equations) and metabotropic (6 equations) receptors. The synaptic time constants associated with the ionotropic receptors are obtained from electrophysiological data while the effective metabotropic time constants are deduced from the associated G protein-coupled inwardly-rectifying potassium (GIRK) current, or if unavailable, the temporal change in firing rate due to the injection of 5-HT or Ox [67]. For simplicity, a simple linear relationship is assumed between the input current I_j,i and activity level f_i. The other important parameter, J_syn,j,i is the connection strength within or between the neuronal populations. Under steady state condition (dI_j,i/dt = 0), J_syn,j,i is defined as the ratio of the current I_j,i and the associated (presynaptic) activity f_i. These currents are obtained from various experiments [4], [26], [27], [31], [68], [69]. The in vivo baseline firing rates (f_i) for the Ox and GABAergic populations in LHA are ∼5 Hz (3–8 Hz in experiments) [70], [71], [72]. In DRN, the baseline neuronal firing rate of 5-HT and GABAergic neuronal population is ∼5 and ∼15 Hz, respectively [73]. The relationship among these baseline activities will be used to constrain our model parameters, namely the J_i, g_i and I_j,0 (Table 1). Consistent with our assumption on ignoring the relatively much faster neuronal membrane dynamics (∼10 s ms), we shall also ignore the dynamics for the relatively fast GABAergic synapses (∼4 ms), assuming they attain instantaneous steady states. This further reduces 4 dynamical equations in describing the associated currents.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Time constants, currents of all the neuronal groups, values are deduced from experiments, and by using above constraints and Connection strength (C. strength), and background current of all the neuronal groups.

https://doi.org/10.1371/journal.pone.0088003.t001

Thus the free parameters in the model are: J_{GABA(DRN)-to-GABA(DRN)}, J_{GABA(LHA)-to-GABA(LHA)} connection strengths of the GABAergic self-inhibition in DRN and LHA neuronal groups; J_{5-HT-to-GABA(LHA)} connection strengths of the effect of 5-HT on GABAergic neurons in LHA; τ _{5-HT-on-GABA(LHA)} time constants of the 5-HT effects on GABAergic neurons in LHA; τ_{Ox-on-GABA(LHA)} time constant of the effect of Ox on GABAergic neurons in LHA; and afferent background currents to the neuronal groups, I_{Background-5-HT}, I_{Background-Ox}, I_{Background-GABA(DRN)}, and I_{Background-GABA(LHA)}.

In addition we make the following further constraints on the values of J’s andτ’s:

1. Time constant of 5-HT effect on LHA’s GABAergic neurons is equivalent to that of 5-HT on LHA’s Ox neurons, i.e. τ_{5-HT-on-GABA(LHA)}∼τ_5-HT-on-Ox
2. Time constants of Ox (Ox₂) effect on GABAergic neurons in LHA is equivalent to the time constant of self-excitation of Ox (Ox₂) autoreceptors in LHA, i.e. τ_{Ox-on-GABA(LHA)}∼ τ_Ox-auto
3. Connection strengths of 5-HT on GABAergic neurons (LHA) are equivalent to the connection strengths of 5-HT on Ox neurons in LHA, i.e. J_{5-HT-to-GABA(LHA)}∼J_5-HT-to-Ox
4. Connection strength of Ox on GABAergic neurons (LHA) is equivalent to the connection strength of the Ox autoreceptors in LHA, i.e. J_{Ox-to-GABA(LHA)}∼J_Ox-auto
5. Connection strengths of GABAergic (GABA_A) self-inhibition in DRN or LHA is equivalent, i.e. J_{GABA(LHA)-to-GABA(LHA)}∼J_{GABA(DRN)-to-GABA(DRN)}.

Further details of the model parameter values, justifications, and their related references are summarized in Table 1.

5-HT_3AR, 5-HT_1BR, 5-HT_2AR and 5-HT_2CR Receptors on Ox Neurons in the LHA

Muraki et al. (2004) have shown the presence of 5-HT_1A receptors on Ox neurons in the LHA. To elucidate the presence of additional 5-HT receptor subtypes on Ox neurons, we performed single IF staining for 5-HT_1BR, 5-HT_2AR and 5-HT_2CR on Ox-EGFP transgenic mice brains (Figure 1). Co-localization of the receptor and the Ox neuron immunoreactivity (Figure 1A–C) can be observed in the respective overlay images. Using double IF labeling, we found fast ligand-based 5-HT_3A receptors on Ox neurons in the LHA of wild type C57BL/6 mice (Figure 2). This suggests that the DRN-to-LHA connection may transmit signals fast. Figure 3 shows representative z-stack_max (30 µm thickness) confocal images showing colocalization of Orexin A with 5HT_1AR (Figure 3A) and 5HT_3AR (Figure 3B), respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Single-label immunofluorescence analysis showing Ox neurons expressing additional 5HT receptor subtypes.

(A) 5- HT_1BR, (B) 5-HT_2AR and (C) 5-HT_2CR in the LHA of EGFP transgenic mice expressing GFP (green) exclusively in the Ox neurons. Immunoreactivity for 5-HT receptors (Alexa 546, red), 100× magnification.

https://doi.org/10.1371/journal.pone.0088003.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Double-label immunofluorescence analysis showing 5-HT_3AR (anti-HTR3A) immunoreactivity of Ox (Orexin-A) neurons in the LHA.

Confocal microscopy: Chromogens were Alexa 488 (green) and Alexa 546 (red), 100× magnification. (A) Arrows indicate the presence of 5-HT_3AR on Ox neurons in the LHA. (B) Enlarged image of the selected square in A.

https://doi.org/10.1371/journal.pone.0088003.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Representative z-stack_max (30 µm thickness) confocal images.

Colocalization of Orexin A with (A) 5HT_1AR, (B) 5HT_3AR, 63× magnification. White arrows indicate orexin neuron expressing the respective 5HT receptor and white boxes indicate orexin neuron not expressing the receptor.

https://doi.org/10.1371/journal.pone.0088003.g003

As we demonstrated the additional 5HT receptor subtypes on the Ox neurons in LHA for the first time, the next obvious step was to quantify the double-labeled Ox neurons for each 5HT receptor subtype. Using the JACoP plug-in tool, we quantified the degree of co-localization by calculating Pearson’s correlation coefficient (Figure 4B) and overlap coefficient by Manders (Figure 4C). It is important to note, that the two coefficients, namely Pearson’s Correlation Coefficient and Overlap Coefficient, showed similar pattern of changes while revealing the different aspects of the colocalization process, proving the applicability of the calculations to investigate the degree of colocalization of serotonin receptor subtypes and orexin A (Ox neurons).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Quantitative analysis of the degree of colocalization of serotonin receptor subtypes and Ox neurons in the LHA.

(A) Representative Scatter plot (cytofluorogram) indicating colocalization of 5HT_1AR and Orexin A, Pearson’s coefficient = 0.641, (B) Pearson’s correlation coefficient calculated for the estimation of degree of colocalization of serotonin reseptor subtypes (5HT_1AR, 5HT_3AR, 5HT_1BR, 5HT_2AR and 5HT_2CR) and Ox neurons in the LHA, (C) Overlap coefficient by Manders calculated for the estimation of extent of overlap of serotonin receptor subtypes (5HT_1AR, 5HT_3AR, 5HT_1BR, 5HT_2AR and 5HT_2CR) and Ox neurons in the LHA. Data was analysed by two tailed unpaired t-test (Prism 5 software). Level of significance ****(p<0.0001).

https://doi.org/10.1371/journal.pone.0088003.g004

5-HT_3AR and 5-HT_1AR Receptors on Ox as well as GABAergic Interneurons in the LHA

To examine the serotonergic long-range connections from DRN in the midbrain to the Ox and GABAergic neurons in the LHA, triple-label IF staining was performed. We searched for 5-HT_3AR (Figure 5A) and 5-HT_1AR (Figure 5B) on the Ox and GABAergic neurons and found that these 5-HT receptors may be present on Ox as well as local GABAergic neurons. This may suggest both direct and indirect (via LHA’s GABAergic neurons) 5-HT influence on the Ox neurons.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Representative photomicrographs of triple-label IF staining in LHA for Ox (Orexin-A) and GABAergic (anti-GABA) neurons.

(A) 5-HT_3AR, overlay image indicates an overlay of Orexin-A, 5-HT_3AR and anti-GABA, (B) 5-HT_1AR, overlay image indicates an overlay of Orexin-A, 5-HT_1AR and anti-GABA. Confocal microscopy: Chromogens were Alexa 488 (blue), Alexa 546 (red) and CF633 (green), 40× magnification.

https://doi.org/10.1371/journal.pone.0088003.g005

OX₁R, OX₂R and NMDAR1 Receptors on Ox Neurons but not on GABAergic Neurons in the LHA

To identify the role of Ox neurons in regulating the activity of local GABAergic neurons, we carried out a triple IF labeling in the LHA. To study this inter-relationship, we searched for OX₁R and OX₂R receptors on the Ox and GABAergic neurons. We identified OX₁R (Figure 6A) and OX₂R (Figure 6B) on the Ox neurons but found these to be absent on the GABAergic interneurons. This sheds light on the self-regulatory mechanism of Ox neurons via OX₁R and OX₂R auto-receptors. Previous studies have demonstrated the presence of inhibitory GABA_B receptors on the Ox neurons [33], [75], [76], [77] and there is evidence that GABA_A receptors are also present on Ox neurons (Backberg et al., 2004, Kokare et al., 2006). This could suggest a one-way relationship between GABAergic and Ox neurons wherein GABAergic neurons exerts an inhibitory effect on the Ox neurons under the partial DRN’s serotonergic control, consistent with Yamanaka et al. (2010).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Representative photomicrographs of triple-label IF staining for Ox and GABAergic neurons in LHA.

(A) OX₁R, overlay image indicates an overlay of Orexin-A, OX₁R and anti-GABA, (B) OX₂R, overlay image indicates an overlay of Orexin-A, OX₂R and anti-GABA. Arrows in A and B overlay images indicate presence of OX₁R and OX₂R on Ox neurons, respectively. Confocal microscopy: Chromogens were Alexa 488 (blue), Alexa 546 (red) and CF633 (green), 40×.

https://doi.org/10.1371/journal.pone.0088003.g006

Ox neurons form a distinct group of a hypothalamic neuronal population that project to multiple brain regions and coordinate many physiological functions [9], [78]. Also, there is a convergence of signals that regulate the activity of Ox neurons by neurotransmitters, hormones, etc. Glutamate is an important neurotransmitter for Ox neurons and excitatory AMPA receptors have been shown to mediate the miniature EPSC in Ox neurons [79]. Also, NMDA receptors activate Ox neurons in the perifornical region of LHA [80]. To verify whether NMDA receptors are present on the overall Ox neurons and not confined solely to the perifornical area, we did triple-label IF staining. Here, we show the presence of N-methyl D-aspartate receptor 1 (NMDAR1) on the Ox neurons (Figure 7) and their absence on the GABAergic interneurons (Figure 7A, overlay image).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. Representative photomicrographs of triple-label IF staining for NMDAR1 on Ox (Orexin-A) and GABAergic (anti-GABA) neurons.

(A) NMDAR1 (anti-NMDAR1) in the LHA. (B) 100× magnification in A. Magenta colour in the overlay image in A and yellow/orange in the overlay image in B indicates the presence of NMDAR1 on Ox neurons. Confocal microscopy: Chromogenes were Alexa 488 (blue in A and green in B), Alexa 546 (red) and CF633 (green), 40×.

https://doi.org/10.1371/journal.pone.0088003.g007

Ox Axonal Projections Receive Glutamatergic and GABAergic Post-synaptic Inputs in the DRN

To study the long-range connections of Ox neurons in the DRN, we performed IF labeling in the DRN region for PSD-95 (a marker for glutamatergic synapses) and gephyrin (a marker for GABAergic synapses) post-synaptic proteins [81]. Firstly, we carried out double-label IF staining for PSD-95 and Ox in the DRN and observed PSD-95 immunopositive Ox axonal terminals (Figure 8) indicating glutamatergic input to the Ox axonal projections. In another set of experiments, we examined GABAergic post-synaptic input to the Ox terminals as it has been shown that Ox fibres project to both 5-HT and GABAergic neurons in the DRN [31]. A double-label IF analysis for gephyrin and Ox in the DRN (Figure 9) showed gephyrin immunopositive Ox terminals, further indicating GABAergic input to the Ox projections. This result is in agreement with the findings of Liu et al. (2002) where they showed Ox axons in proximity to GABA/GABA-transporter immunoreactive neurons.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 8. Ox axonal projections receive glutamatergic post-synaptic inputs in the DRN.

(A) Ox (Orexin-A) axonal projections in the DRN, Aq indicates aqueduct, 20× magnification. (B) Double IF staining was done for Ox axons and PSD-95 (post-synaptic glutamatergic marker) in the DRN, 100×. Confocal microscopy: magnification 100×. (C) Z-stack_max image of 30 µm thickness selected from 45 µm section.

https://doi.org/10.1371/journal.pone.0088003.g008

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 9. Ox axonal projections receive GABAergic post-synaptic inputs in the DRN.

(A) Double IF staining was performed for Ox axons (Orexin-A) and Gephyrin (post-synaptic GABAergic marker) in the DRN. Confocal microscopy: magnification 100×. (B) Z-stack_max image of 30 µm thickness selected from 45 µm section.

https://doi.org/10.1371/journal.pone.0088003.g009

Map of the DRN-LHA Circuit

Based on the above results and in other previous studies (see below), a tentative map of the DRN and LHA is illustrated in Figure 10A. Connection (i) in the figure involves 5-HT_1AR [26], 5-HT_1BR, 5-HT_2AR, 5-HT_2CR, 5-HT_3AR (found in our current study). Although connection (ii), found in this study, consists of 5-HT_1AR and 5-HT_3AR, the specific types of connection (e.g. effectively excitatory or inhibitory) and their strengths are not known. Connection (iii) and its receptor types are not known yet. Similarly, the receptor types for connection (iv) (Ox₁R and Ox₂R found in this study are consistent with previous findings [30], [32]), and connections (v) and (vi) (found in this study) are unknown. Connection (vii) receptors are not known and specifically Ox₁R and Ox₂R are not found in our study.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 10. Neural circuit map of the LHA-DRN system.

(A) Diagrammatic representation of the receptor mapping suggests a complex bi-directional relationship between DRN and LHA. Black circle/arrow represents effective inhibitory/excitatory connections between LHA and DRN, solid black line represents various receptor types derived from previous experimental studies, dashed black line shows the recent experimental findings, dashed arrows with * (asterisk) indicates that some of these receptors are known in the previous studies, dashed-dotted line signifies that receptors (in the target neurons) are not identified in this current study, dotted lines shows the hypothesized connection, diamond arrows indicate that connection types (excitatory/inhibitory) are not known, ? (question mark) means that receptors are not known and gray line signifies that the connection is not studied. All these receptor types are numbered on the figure and details are provided below. See text for more information. (B) Reduced LHA-DRN circuit used in the computational model. Label as in (A).

https://doi.org/10.1371/journal.pone.0088003.g010

Other connections in the circuit based on other previous work: (viii) excitatory connection and receptor types are not known [82]; (ix) GABA_A/B [33], [75], [76], [77]; (x) AMPAR [79] and NMDAR1 (found in the current study); (xi) GABA_B [77]; (xiii) Ox₂R [4], Ox₁R and Ox₂R (found in the current studies); (xv) 5-HT_1B/1D [34]; 5-HT_1A/2A/2C [69]; (xvi) 5-HT₇ [34]; (xvii) GABA_A/B [34]; (xviii) AMPAR and NMDAR [34]; (xix) AMPAR and NMDAR [34]; (xxi) 5-HT_1AR [83]. For the connections (xii), (xiv), (xx) and (xxii) (shown by black dotted circle/arrow in Figure 10A), we implicitly assume that non-principal neurons (GABAergic and glutamatergic neurons in LHA and DRN) are self-coupled.

Neural Circuit Modelling and Analysis of the Direct and Indirect Interactions within the DRN-LHA System

The experimental findings in this study and in other previous work now provide sufficient information to build a computational model to investigate how the DRN-LHA neural architecture can influence the systems dynamics. We purposely simplified the implementation of the neural population units and connections in order to better illustrate the effects of network topology on dynamics (see Materials and Methods section). We also omit modelling the glutamatergic neurons in the DRN and LHA because there is evidence that glutamate has a weaker effect than GABA in the DRN [59]; and that indirect excitation from LHA’s glutamatergic neurons on Ox neurons may be represented implicitly by the Ox autoreceptors. This significantly reduces the number of model parameters and dynamical equations involved. In our model, we find that the various receptor subtypes will not affect the steady-state activities of the system. We shall henceforth not elaborate on the specific receptor subtype till we investigate the transient activation part of the results (section 3.6.3) when we study the influence of various timescales induced by the various receptors.

Next we investigate how the unknown model parameters and different timescales affect the system behaviour. Specifically, the focus is to understand the effects of: (i) 5-HT on LHA’s GABAergic neurons; (ii) local Ox and GABAergic interactions; and (iii) how connection timescales affect phasic 5-HT or Ox activations.

Oscillatory DRN-LHA Behaviour if 5-HT Weakly Excites or Inhibits LHA’s GABAergic Neurons

It is not yet known whether the connection from 5-HT neurons to LHA’s GABAergic neurons (J_{5-HT-to-GABA(LHA)}) is effectively excitatory or inhibitory. Here, we shall explore these possibilities. If this connection is excitatory, we find that as its connection strength J_{5-HT-to-GABA(LHA)} is increased, the steady-state firing rate activities for most of the neural populations decrease (Figure 11). Note that LHA’s GABAergic neural population barely increases. This slight change is due to the self-inhibition within these GABAergic neurons. In contrast, the relatively larger changes for the other neural populations (especially when the connection strength is weaker) are due to the strong inhibitory projection from the LHA’s GABAergic neuron onto the Ox neurons (see Table 1), which subsequently affect the neurons in the DRN.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 11. Neural circuit responses to change in connection strength from 5-HT to GABAergic neurons (LHA).

Change in the steady-state values of the neural firing frequencies of the neuronal groups with varying connection strength factor J_{5-HT-to-GABA(LHA)} (in pA/Hz) from 5-HT neurons to the GABAergic neurons (LHA) for excitatory connection. Steady states are obtained after simulating for a sufficiently long time. f_5-HT, f_Ox, g_GABA(LHA), and g_GABA(DRN): population firing frequencies of 5-HT, Ox, LHA’s GABAergic and DRN’s GABAergic neurons, respectively.

https://doi.org/10.1371/journal.pone.0088003.g011

When the connection from 5-HT to LHA’s GABAergic neurons becomes very weak (J_{5-HT-to-GABA (LHA)} close to 0), the DRN-LHA circuit will begin to exhibit slow oscillatory behaviour. Figure 12 shows such transition towards oscillatory behaviour for one sample neural population (the rest of the neural populations look similar). The oscillatory period can be as slow as a few minutes (Figure 12, inset). It can also be observed that as the connection strength decreases, the amplitude of oscillation (bounded by the top and bottom lines in grey region) increases. When this particular connection becomes inhibitory (J_{5-HT-to-GABA(LHA)} <0), the oscillation amplitude can become very large (not shown). It is well-known that excitatory-inhibitory network can easily create oscillation [60], [61]. Similarly, the observed oscillation phenomenon can be explained by the interplay between strong inhibition and self-excitatory auto-regulation in the Ox neural population; the oscillation disappears when Ox autoreceptors are removed from the model e.g. when J_Ox-auto = 0 (not shown). As far as we know, there has yet to be any observable slow oscillation of Ox in the timescale of minutes. Hence, our model suggests that strong excitatory connection from 5-HT to LHA’s GABAergic neurons is more plausible, and we shall proceed with this assumption from here on.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 12. LHA-DRN system can exhibit oscillations with weak excitatory from 5-HT to GABAergic (LHA) neurons.

Steady-state values of the firing rate activity of 5-HT neurons are plotted as a function of the connection strength J_{5-HT-to-GABA(LHA)} (in pA/Hz). Oscillatory region (left of dashed) is bounded by the values of J_{5-HT-to-GABA(LHA)} below 0.804 pA/Hz. Max (Min): maximum (minimum) firing rates during oscillation. Label as in Figure 11.

https://doi.org/10.1371/journal.pone.0088003.g012

DRN-LHA System is Resilient to Changes in the Ox-to-GABAergic Neurons in the LHA

One of our experimental findings in this study shows that the LHA’s GABAergic neurons do not have Ox receptors, consistent with indirect results from previous works [4]. Using our model, we check for the significance of such Ox receptors’ absence. In our model, we gradually increase the connection strength of Ox to LHA GABAergic neurons (J_{Ox-to-GABA(LHA)}). This only marginally decreases the DRN-LHA steady-state activities (Figure 13). A similar explanation as that for the results in Figure 11 can be used to account for this phenomenon.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 13. Neural circuit responses to change in connection strength from Ox to GABAergic neurons (LHA).

Steady-state firing frequencies of the neural populations as functions of the connection strength (in pA/Hz). Label as in Figure 11.

https://doi.org/10.1371/journal.pone.0088003.g013

Transient Ox and 5-HT Activities can be Affected by Different Receptor Timescales

Having studied the tonic activities (stable steady states) of the LHA-DRN system, we shall now proceed to investigate how transient or phasic activations of 5-HT or Ox can affect the system. The motivation for this is that it has been known that 5-HT and Ox can phasically activate in the presence of behaviourally relevant stimuli [84], [85], [86]. Furthermore, our current experimental finding suggests that 5-HT can influence LHA neurons over multiple timescales, through both the slow G-protein-coupled (non-5-HT_3A) and fast ligand-gated (5-HT_3A) receptors. To simulate this, the (more plausible) excitatory connection J_{5-HT-to-GABA(LHA)} is considered, and a pulse stimulus current of the duration of 0.5 s with an amplitude of 150 pA is applied to either the 5-HT or Ox neural populations. To understand the individual role of the slow and fast timescales, and minimize any confounding effect, we simulate the two timescales separately.

Figure 14A (left) shows that a 0.5 s stimulation of 5-HT neurons rapidly increase its activity, followed by a slower decay back towards baseline. The 5-HT activity decay is due to feedback inhibition from its autoreceptors and the local GABAergic neurons (hence the undershoot below baseline). The Ox activity responses are generally suppressed by the strong 5-HT-mediated inhibition (Figure 14A, right). When the 5-HT-to-Ox and 5-HT-to-GABA (LHA) connections are fast (mimicking 5-HT_3AR), the rebound upon removal of stimulus (∼ after the 2 s mark in Figure 14A, right) is higher, suggesting the faster disinhibition than that for the slower connections. Varying these timescales do not affect the transient 5-HT activity in DRN (overlapping curves, Figure 14A, left).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 14. Transient activities of 5-HT and Ox neuronal groups under phasic stimulation.

(A) Firing frequency of 5-HT neurons for different slow and fast 5-HT receptor timescales (left) and for the Ox neurons (right). (B) Firing frequency of 5-HT neurons for different slow and fast Ox receptor time scales (left) and for the Ox neurons (right). Solid horizontal black lines denote presence of stimulus with duration of 500 ms, and amplitude of 150 pA. Label as in Figure 11.

https://doi.org/10.1371/journal.pone.0088003.g014

When a similar stimulus is applied to the Ox neurons, the activity of the Ox neurons increases considerably throughout the stimulus duration (Figure 14B, right), which is due to the self-amplifying effect of Ox autoreceptors dominating over the local GABAergic inhibition (feedback inhibition is not strong due to the weak excitatory connection from Ox neurons to their local GABAergic neurons). These Ox neurons, affect 5-HT neurons either directly or indirectly (through the DRN’s GABAergic neurons). From our simulations, we find that when the Ox-to-5-HT connection acts on a fast timescale (5 s), they can transiently excite the 5-HT neurons (Figure 14B, left). If the direct timescale is much slower (e.g. 60 s), then 5-HT neurons can hardly be activated. Varying Ox-to-GABA(DRN) connection does not affect the system significantly. Note the post-stimulus undershoot due to recurrent inhibition, suppressing the Ox and 5-HT transiently below baseline. Taken together, when Ox-to-5-HT acts on a faster timescale, the phasic influence of Ox on 5-HT activity is much larger.