1. Computational and Systems Biology
  2. Neuroscience
Download icon

CO2-evoked release of PGE2 modulates sighs and inspiration as demonstrated in brainstem organotypic culture

  1. David Forsberg
  2. Zachi Horn
  3. Evangelia Tserga
  4. Erik Smedler
  5. Gilad Silberberg
  6. Yuri Shvarev
  7. Kai Kaila
  8. Per Uhlén
  9. Eric Herlenius  Is a corresponding author
  1. Karolinska Institutet, Sweden
  2. Karolinska University Hospital, Sweden
  3. University of Helsinki, Finland
Research Article
  • Cited 8
  • Views 1,964
  • Annotations
Cite this article as: eLife 2016;5:e14170 doi: 10.7554/eLife.14170

Abstract

Inflammation-induced release of prostaglandin E2 (PGE2) changes breathing patterns and the response to CO2 levels. This may have fatal consequences in newborn babies and result in sudden infant death. To elucidate the underlying mechanisms, we present a novel breathing brainstem organotypic culture that generates rhythmic neural network and motor activity for 3 weeks. We show that increased CO2 elicits a gap junction-dependent release of PGE2. This alters neural network activity in the preBötzinger rhythm-generating complex and in the chemosensitive brainstem respiratory regions, thereby increasing sigh frequency and the depth of inspiration. We used mice lacking eicosanoid prostanoid 3 receptors (EP3R), breathing brainstem organotypic slices and optogenetic inhibition of EP3R+/+ cells to demonstrate that the EP3R is important for the ventilatory response to hypercapnia. Our study identifies a novel pathway linking the inflammatory and respiratory systems, with implications for inspiration and sighs throughout life, and the ability to autoresuscitate when breathing fails.

https://doi.org/10.7554/eLife.14170.001

eLife digest

Humans and other mammals breathe air to absorb oxygen into the body and to remove carbon dioxide. We know that in a part of the brain called the brainstem, several regions work together to create breaths, but it is not clear precisely how this works. These regions adjust our breathing to the demands placed on the body by different activities, such as sleeping or exercising. Sometimes, especially in newborn babies, the brainstem’s monitoring of oxygen and carbon dioxide does not work properly, which can lead to abnormal breathing and possibly death.

In the brain, cells called neurons form networks that can rapidly transfer information via electrical signals. Here, Forsberg et al. investigated the neural networks in the brainstem that generate and control breathing in mice. They used slices of mouse brainstem that had been kept alive in a dish in the laboratory. The slice contained an arrangement of neurons and supporting cells that allowed it to continue to produce patterns of electrical activity that are associated with breathing. Over a three-week period, Forsberg et al. monitored the activity of the cells and calculated how they were connected to each other. The experiments show that the neurons responsible for breathing were organized in a “small-world” network, in which the neurons are connected to each other directly or via small numbers of other neurons.

Further experiments tested how various factors affect the behavior of the network. For example, carbon dioxide triggered the release of a small molecule called prostaglandin E2 from cells. This molecule is known to play a role in inflammation and fever. However, in the carbon dioxide sensing region of the brainstem it acted as a signaling molecule that increased activity. Therefore, inflammation could interfere with the body’s normal response to carbon dioxide and lead to potentially life-threatening breathing problems. Furthermore, prostaglandin E2 induced deeper breaths known as sighs, which may be vital for newborn babies to be able to take their first deep breaths of life. Future challenges include understanding how the brainstem neural networks generate breathing and translate this knowledge to improve the treatment of breathing difficulties in babies.

https://doi.org/10.7554/eLife.14170.002

Introduction

Breathing is essential for life, but the underlying mechanisms that control breathing movements and neuronal pattern generation are under debate (Jasinski et al., 2013). Breathing maintains tissue homeostasis, and an adequate response to increased carbon dioxide (CO2) levels is crucial (Kaila and Ransom, 1998; Guyenet and Bayliss, 2015). Failure to adequately respond to pCO2 alterations is linked to breathing disturbances; apnea of prematurity; centrally mediated sickness, such as noxious sensations and panic; and premature death, as in sudden infant death syndrome (Guyenet and Bayliss, 2015).

Neuronal networks in the parafacial respiratory group/retrotrapezoid nucleus (pFRG/RTN) and the preBötzinger complex (preBötC) are important networks implicated in the central control of breathing. pFRG/RTN paired-like homeobox 2b (Phox2b)-expressing neurons are sensitive to changes in CO2 levels or their proxy, pH ([H+]) (Mellen and Thoby-Brisson, 2012; Onimaru and Dutschmann, 2012). This responsiveness to hypercapnia is independent of synaptic transmission, and the Phox2b+ neurons detect CO2/H+ via intrinsic proton receptors (TASK-2 and GPR4) in parallel pathways (Kumar et al., 2015). Moreover, medullary astrocytes contribute to central chemosensitivity. Slight acidification leads to an increased astrocytic intracellular concentration of calcium ions (Ca2+), resulting in vesicle-independent ATP release (Gourine et al., 2010).

In addition, a CO2 sensitivity of astrocytes also mediates a vesicular-independent ATP release (Huckstepp and Dale, 2011). Some connexins, which are expressed on astrocytes, e.g., connexin 26 (Cx26) and Cx30, are indeed sensitive to CO2 (Meigh et al., 2013; Reyes et al., 2014).

These cellular processes of chemosensitivity result in an altered respiratory pattern that lowers the blood CO2 levels. Inflammation reduces the CO2 response and, particularly in neonatal mammals, can induce sighs, an altered response to hypoxia and potentially life-threatening apnea episodes as shown in humans, sheep, piglets and rodents (Guerra et al., 1988; Long, 1988; Herlenius, 2011; Siljehav et al., 2014; Koch et al., 2015; Siljehav et al., 2015).

In the inflammatory pathway, prostaglandin E2 (PGE2) is an important molecular mediator, that together with its main receptor, the EP3R, play roles in the hypoxic and hypercapnic responses, e.g. seen in patients with bronchopulmonary dysplasia (Kovesi et al., 2006; Siljehav et al., 2014; Koch et al., 2015). PGE2 also seems to induce a sigh oriented respiratory pattern (Koch et al., 2015). Sighs are regularly occurring events of augmented breaths with a biphasic inspiratory pattern with the initial phase being comparable to eupnea and the second having larger amplitude (Toporikova et al., 2015). Such breaths are necessary for life and have been linked to several pathological states (Ramirez, 2014; Li et al., 2016).

Here, we hypothesized that both PGE2 and EP3R constitute parts of the respiratory machinery and that they are involved in the induction of sighs and the hypercapnic response. We established a viable brainstem organotypic slice culture that maintains respiratory-related activity for several weeks in vitro and used this to investigate how PGE2 and EP3R alter breathing and control of chemosensitivity. Our novel data reveal an important role of the EP3R in the pFRG/RTN hypercapnic response and furthermore suggest that PGE2 is released during hypercapnia, possibly through CO2-sensitive connexin hemichannels. Inflammation, with its associated PGE2 release, exogenous PGE2 and a lack of EP3R, blunts the hypercapnic response. These data link the inflammatory and respiratory systems, with implications for sighs and inspiration throughout life as well as for the ability to autoresuscitate when breathing fails.

Results

EP3R is involved in respiratory control, sighs and the hypercapnic response

To investigate the role of PGE2 and EP3R in respiration and sigh activity, we performed whole body plethysmography on 9-day old mice. We found EP3R and its ligand PGE2 to be important modulators of breathing and the response to hypercapnia (5% CO2 in normoxia; Table 1). The sigh frequency increased after the intracerebroventricular (i.c.v.) injection of PGE2 (1 µM in 2–4 µl artificial cerebrospinal fluid, aCSF) in an EP3R-dependent manner (Figure 1c–d, Table 2), as did the tidal volume (VT) (during eupnea, excluding sighs) in wild-type mice (Figure 1e). Furthermore, hypercapnic exposure also induced an increase in sigh frequency (Figure 1f, Table 2). This increase was larger in wild-type mice than in mice lacking the EP3R (Ptger3-/- mice). This CO2-induced increase in sigh frequency was abolished in wild-type mice after i.c.v. injection of PGE2 (Figure 1f, Table 2). The mice also responded to hypercapnia with increases in respiratory frequency (FR), VT and minute ventilation (VE; Figure 1g). I.c.v. injection of PGE2 abolished the VT but not the FR response during hypercapnia (Table 1). This provides new information on how PGE2 induces sigh activity and how increased PGE2 levels, as during inflammation, may both induce sighs and attenuate responsiveness to CO2.

Table 1

Respiratory parameters under basal conditions. Ptger3-/- mice are heavier than wild-type mice of the same age. They do not, however, differ in respiratory frequency (FR), tidal volume (VT), or minute ventilation (VE). I.c.v. injection of PGE2 increases VT and VE in wild-type mice but not Ptger3-/- mice. Respiratory frequency, tidal volume, and minute ventilation all increased during hypercapnic exposure. n: number of animals. Data are presented as mean ± SD.

https://doi.org/10.7554/eLife.14170.003
Weight (g)F(breaths/min)V(µl /g)V(µl/g/min)F(breaths/min)
Hypercapnia
V(µl /g)
Hypercapnia
V(µl /g)
Hypercapnia
WT - vehicle
n=5
3.7 ± 0,5#206 ± 28*9.7 ± 2.9* †2.0 ± 0.8*†259 ± 20*#12.3 ± 3.1*#3.2 ± 1.0*#
WT - PGE2
n=8
3.9 ± 0.4210 ± 15*15.1 ± 3.3†3.2 ± 0.8*†267 ± 31*#15.9 ± 2.64.2 ± 0.8*#
Ptger3-/- - vehicle
n=5
4.8 ± 0.4#215 ± 32*14.2 ± 2.4*3.0 ± 0.5*240 ± 37*#15.7 ± 3.1*#3.4 ± 1.1*#
Ptger3-/- - PGE2
n=7
4.4 ± 0.3211 ± 18*13.8 ± 2.7*2.9 ± 0.6*241 ± 30*#14.9 ± 2.7*#3.5 ± 0.4*#
  1. *p<0.05 (normocapnia vs. hypercapnia),

  2. #p<0.05 (WT vs. Ptger3-/-),

  3. †p<0.05 (vehicle vs. PGE2).

PGE2 and CO2 increase sigh activity via EP3R signaling.

Respiratory activity was recorded in vivo in a two-chamber plethysmograph (a). Sighs, defined by an increase in inspiratory volume and respiratory cycle period with a biphasic inspiration (b), increase in frequency after intracerebroventricular injection (i.c.v.) of PGE2. This effect is absent in mice lacking EP3R (Ptger3-/-, c, arrows, d). I.c.v. injection of PGE2 also increases the tidal volume (VT) in wild-type C57BL/6J (WT) mice (e). The sigh frequency is increased by hypercapnic (5% CO2 in normoxia) conditions in wild-type and Ptger3-/- mice but less so in Ptger3-/- mice (f). In wild-type mice, the increase is abolished after i.c.v. injection of PGE2 (f). Hypercapnic exposure causes an increase in respiratory frequency (FR), tidal volume (VT), and minute ventilation (VE) (g), but the increase is attenuated in Ptger3-/- mice. Data are presented as means ± SD. *p<0.05 Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.004
Table 2

PGE2 and hypercapnia induce sighs. Sigh frequency does not differ between wild-type mice and Ptger3-/- mice. In wild-type mice, PGE2 increases sigh frequency. Hypercapnia also increases sigh frequency more in wild-type mice than in Ptger3-/- mice. PGE2 abolishes this increase in wild-type mice but not in Ptger3-/- mice (*p<0.05). n: number of animals. Data are presented as mean ± SD.

https://doi.org/10.7554/eLife.14170.006
Sighs/min NormocapniaSighs/min Hypercapnia
WT - vehicle n=50.4 ± 0.11.0 ± 0.2*
WT - PGE2 n=80.7 ± 0.1*0.9 ± 0.3
Ptger3-/-- vehicle n=50.4 ± 0.10.7 ± 0.3*
Ptger3-/—- PGE2 n=70.4 ± 0.10.6 ± 0.2*

To unravel the mechanistic details of the PGE2-EP3R system in respiratory regulation and its connection to the hypercapnic response and sighs, we set out to create a model system that would allow long-term, detailed studies of the respiratory neural networks, i.e., networks with neurons as well as glial cells.

Establishment of a viable respiratory brainstem organotypic slice culture

Brainstem organotypic slice cultures of the mouse brainstem from 3-day-old mice were prepared at the preBötC brainstem level (Figure 2a). To validate this new model system, we first examined survival and expression of various neural markers in the brainstem slice cultures during cultivation.

Neural marker staining showed intact neurons, and neurokinin 1 receptor (NK1R)-positive respiratory regions were cytoarchitectonically well preserved (Figure 2b,e,g, Figure 2—figure supplement 1). The expression pattern of vesicular glutamate transporter 2 (VGlut2), similar to that in vivo, indicates the functional potential of the brainstem slice culture because glutamatergic synapses are essential for the development of the breathing rhythm generator (Wallén-Mackenzie et al., 2006) (Figure 2d). Neuronal markers MAP2 and KCC2 (Kaila et al., 2014) were expressed in the preBötC (Figure 2c–f, Figure 2—figure supplement 2). The protein expression in the preBötC remained stable for 3 weeks of cultivation (Figure 2—figure supplement 1). The brainstem slice cultures became thinner with longer cultivation as the tissue spread out (Figure 2—figure supplement 2). However, they remained viable and exhibited a low degree of necrosis and apoptosis, even after 3 weeks (Figure 2—figure supplement 3).

Figure 2 with 3 supplements see all
Brainstem slice cultures have a preserved structure and neurons with functional potential.

Brainstem slices containing the preBötC were used to create slice cultures. Anatomical landmarks, including the nucleus ambiguus (NA), nucleus tractus solitarius (NTS), and nucleus hypoglossus (XII; a), as well as the distinct expression of NK1R (b, c, g) enabled the identification of the preBötC region. The brainstem slice displayed MAP2-/Tuj1-positive neurons expressing NK1R (bc), VGlut2 (d), and/or KCC2 (e). The abundant MAP2-/Tuj1-positive cells demonstrated a preserved neuronal network within the preBötC (g). KCC2 expression was found in the NTS, NA, and preBötC (e). DIV; days in vitro. Arrowheads: double-labeled cells. Scale bars: 100 µm in b–f, 500 µm in g.

https://doi.org/10.7554/eLife.14170.007

Physiological measurements of brainstem respiratory activity demonstrate functional and responsive networks

After evaluating morphology, we investigated the cellular activity within the brainstem slice culture.

Neurons in the brainstem slice cultures retained their electrical properties at 7 days in vitro (DIV), including a resting membrane potential of −55 ± 6 mV (Figure 3b–c) and overshooting action potentials (Figure 3c). The resting membrane potential, action potential threshold, half-width and peak amplitudes of the action potential, and membrane time constant were within the ranges of acute respiratory slices (Figure 3c, Figure 3—figure supplement 1). Action potentials occurred in clusters of regular rhythmic bursting activity. Neuronal connections were also similar to those seen immediately ex vivo, e.g., in acute slices, (Ballanyi and Ruangkittisakul, 2009) as evidenced by the postsynaptic potentials and concurrent inputs to neighboring neurons, resulting in correlated activity (Figure 3b, Figure 3—figure supplement 1).

Figure 3 with 1 supplement see all
Neuronal electrical activity indicates preserved networks.

Neurons in a preBötC slice (7 DIV), patched in the whole-cell configuration in current-clamp mode (a), exhibit regular rhythmic bursting activity (b). The neurons exhibited a hyperpolarized resting potential, action potentials, synaptic input, and spontaneous electrical activity, with epochs of action potential activity (b, c). The different measured variables indicated healthy and normally functioning neurons (d). Depicted here are two simultaneously patched neurons that also received common synaptic input (e, arrows). Spiking epochs occurred simultaneously, suggesting synchronized network oscillations. Direct connectivity between the depicted neurons showed that they were neither chemically nor electrically synaptically connected to each other. This finding indicates that the observed correlation was induced by common input from a preserved network structure. AP: action potential. DIV: days in vitro. Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.011

Thus, on an individual neuronal level, the cells behave as expected. However, breathing is generated through cellular interactions in respiration-related neural networks.

To investigate how individual cells interact, we applied live time-lapse Ca2+ imaging to allow simultaneous recording of the activity of hundreds of cells. Tetramethyl rhodamine coupled Substance P (TMR-SP), visualizing NK1R-expressing neurons, was used to identify the preBötC. In the brainstem slice cultures, the preBötC contained networks with correlated activity between cells (Figure 4b–d), which was analyzed using a recently reported cross-correlation analysis method (Smedler et al., 2014) (Figure 4—figure supplement 1). We found clusters of cells with highly correlated activity. Such groups of cells in close proximity to each other were interconnected via a few cells that seem to function as hubs (Watts and Strogatz, 1998). The correlated network activity in the preBötC was preserved for 1, 2 and 3 weeks (Figure 4b–e). The number of active cells and the correlations per active cell remained similar over time (Figure 4e). These data suggest that the brainstem slice culture approach can indeed be used to perform long-term studies of respiratory neural network activity.

Figure 4 with 2 supplements see all
Neural activity in the preBötC is arranged in a functional respiratory network with respiratory-related motor output.

In the preBötC slice (a), a cross-correlation analysis of Ca2+ time-lapse imaging data (Figure 4—figure supplement 1) revealed small-world network-structured correlated activity in the preBötC (b–d). The number of correlating cell pairs did not change over time (e), nor did the small-world network parameter or connectivity (f). TMR-SP-positive regions contained more correlated cell pairs than TMR-SP-negative regions (621 ± 284, N=14 and 56 ± 48, N=9, respectively; p<0.05), although there was no difference in the number of active cells (112 ± 57, N=14 and 144 ± 68, N=9, respectively, N.S.; g). As in the preBötC, the nucleus hypoglossus maintained correlated neural network activity (h). Ten percent of the cells (n=8–12/slice) in the hypoglossal nucleus exhibited a regular spiking frequency of ~50–100 mHz (i). The multicolored bar indicates the correlation coefficient in bh; warmer colors indicate more strongly correlated activity between two cells connected by the line. DIV: days in vitro. A.U.: arbitrary units. w: week. N: number of slices, n: number of cells. Scale bars: 500 µm in a, 100 µm in bd and gh. Multicolored bar: color-coded correlation coefficient values. Data are presented as means ± SD. Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.014

Analysis of the network structure revealed stable connectivity values (i.e., the number of cell pairs with a correlation coefficient exceeding the cut-off value, divided by the total number of cell pairs) during the cultivation of preBötC slices for up to 3 weeks (Figure 4f, Table 3). These values were slightly higher than those estimated in a previous study (Hartelt et al., 2008), in which only neurons were accounted for. However, both neurons and glia are involved in respiratory control (Erlichman et al., 2010; Giaume et al., 2010), and our analysis provides information on both cell types. Moreover, other analyzed network parameters, i.e., the normalized mean path-length (λ) and the normalized mean clustering-coefficient (σ), also remained stable (Figure 4f, Table 3). Overall, the small-world parameter (Watts and Strogatz, 1998γ=σλ was unchanged after 3 weeks in culture. Inhibiting the firing of action potentials and consequent activation of synapses by tetrodotoxin (TTX, 20 nM) abolished the coordinated network activity and revealed a population of cells that retained rhythmic alterations of cytosolic Ca2+ levels (31 ± 4% of the total number of cells, N=14 slices). Most of these cells (76 ± 12%, N=14) were NK1R-positive neurons, indicating the presence of functioning pacemaker neurons (Figure 4—figure supplement 2). The Ca2+ signals from synapse-independent cells remained, however with a lower frequency and higher coefficient of variation (Figure 4—figure supplement 2). Regions outside the brainstem nuclei contained active cells, without intercellular coordination (Figure 4g). This cellular activity ceased during TTX treatment. In conclusion, the brainstem slice cultures contain a preserved preBötC network with a small-world structure.

Table 3

The preBötC network parameters remain unchanged for 21-DIV cultures.The results of correlation analysis for the preBötC are shown. N.S.: not significant. N: number of slices. Data are presented as mean ± SD.

https://doi.org/10.7554/eLife.14170.019
preBötC7 DIV (N=12)14 DIV (N=13)21 DIV (N=8)
Correlating cell pairs560 ± 325501 ± 277517 ± 327N.S.
Active cells110 ± 40100 ± 59110 ± 69N.S.
Correlations per active cell6 ± 46 ± 57 ± 6N.S.
Connectivity0.8 ± 0.10.8 ± 0.20.8 ± 0.2N.S.
Mean shortest path length (λ)0.7 ± 0.20.8 ± 0.20.8 ± 0.2N.S.
Clustering coefficient (σ)2.7 ± 1.52.8 ± 1.62 ± 0.6N.S.
Small-world parameter (γ)4.2 ± 3.03.4 ± 1.72.7 ± 1.7N.S.

As the preBötC delivers part of its motor output through the hypoglossal nerve (Smith et al., 2009), we also examined the hypoglossal motor nucleus. In this region of the hypoglossal motor nucleus, we found correlated cell activity organized similarly to that found in the preBötC network (Figure 4h). Within this network, frequency analysis revealed regularly spiking cells with a frequency between 50 and 100 mHz, corresponding to a rhythmic motor neuron output of 3–6 bursts of respiration-related activity/min (average 3.7 ± 0.9 bursts/min; Figure 4i). This suggests a preserved respiratory-related output in the brainstem slice cultures.

Subsequent recordings of extracellular potentials from the 12th cranial nerve and hypoglossal nucleus revealed a corresponding rhythmic respiratory-related output at 7 (N=16), 14 (N=3), and 21 DIV (N=6). Respiratory output from acute slices varied between 1 and 8 bursts per min (neonatal mice, 3 mM K+), with frequencies in the lower range after a longer incubation time in vitro (Ramirez et al., 1997; Ruangkittisakul et al., 2011). In our model we observed a respiratory-related frequency of 3.7 ± 2.5 bursts per min (average of frequencies at 7, 14 and 21 DIV, no significant difference was observed between different DIV, Figure 5a), which is within the expected range for a slice. Among individual cultures, there was some variability in frequency (Figure 5a). However, the intrinsic rhythm was stable, with an average coefficient of variation of 22 ± 8 (no difference between the different DIV, Figure 5b). Rhythmic XII activity was observed for more than 2 hr during recordings (Figure 5—figure supplement 1).The activity could be inhibited by a µ-opioid receptor agonist, [D-Ala2, N-Me-Phe4, Gly5-ol]-enkephalin (DAMGO, 0.5 µM; Figure 5c, Figure 5—figure supplement 1) and stimulated by NK1R agonist Substance P (1 µM; 19 ± 13% increase in frequency, p<0.05; N=7; Figure 5—figure supplement 1).

Figure 5 with 1 supplement see all
Breathing brainstem in a dish: ongoing/persistent rhythmic XII motor activity.

The connected preBötC neural networks generate respiratory-related motor neuronal output delivered through the 12th cranial nerve (XII). The hypoglossal nucleus/nerve discharge frequency varied among the brainstem slice cultures but did not depend on brainstem slice culture age (a, N=16 at 7 DIV, N=3 at 14 DIV, and N=6 at 21 DIV). The regularity of respiration-related motor activity, measured as CV (coefficient of variation), remained stable during 3 weeks of culture (b). The µ-opioid receptor agonist DAMGO (0.5 µM) silenced the XII nerve activity in 5/5 brainstem slice cultures, as depicted here in (c) from a 7-DIV brainstem culture (filtered trace, above, and rectified and smoothed trace, below). DAMGO lowered the Ca2+. In the hypoglossal nucleus, DAMGO (0.5 μM) lowered the frequency of regularly-spiking cells (f, g). N: number of slices. Data are presented as means ± SD. *p<0.05 Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.020

In the preBötC, DAMGO also inhibited the Ca2+ activity of individual NK1R+ neurons and lowered the network frequency significantly (Figure 5d, Video 1). This was accompanied by an increase in the coefficient of variation in this area (36 ± 4 vs. 47 ± 6, N=7 slices, p<0.05). The network structure was not affected. An increase in [K+] from 3 mM to 9 mM, with subsequent membrane potential depolarization, increased the frequency in the preBötC (Figure 5e). In the hypoglossal nucleus, DAMGO caused a frequency reduction in the regularly spiking cells (Figure 5f,g). Thus, the preBötC brainstem slice culture remained active and responsive and generated rhythmic respiration-related motor output activity.

Video 1
NK1R+ respiratory neurons in the preBötC are identified using TMR-SP (red dye), followed by Ca2+ oscillations visualized with Fluo-4.

After 25 s, the µ-opioid receptor agonist DAMGO (0.5 µM) is added and reduces the signaling frequency of the network. fps: frames per second.

https://doi.org/10.7554/eLife.14170.026

Gap junctions are essential parts of correlated preBötC activity

Gap junction signaling plays an important role in the development of the respiratory system, the maintenance of respiratory output and likely the CO2/pH response (Elsen et al., 2008; Fortin and Thoby-Brisson, 2009; Gourine et al., 2010; Huckstepp et al., 2010a). Thus, we used the brainstem slice cultures to investigate the involvement of gap junctions in the neural networks and their response to CO2.

In the brainstem slice cultures, immunohistochemistry showed high Cx43 expression in neurons of the preBötC (Figure 6a) and lower and persistent Cx26 and Cx32 expression in the respiratory regions (Figure 6b–d) at 7 DIV. To assess the function of these intercellular gap junctions and hemichannels, we treated the brainstem slice cultures at 7 DIV with gap junction inhibitors carbenoxolone (CBX) or 18α-glycyrrhetinic acid (18-α-GA). Both inhibitors decreased the number of correlating cell pairs and active cells in the preBötC, whereas glycyrrhizic acid (GZA), an analog to CBX that lacks the ability to block gap junctions, and the aCSF control did not (Figure 6e–g, k–l). However, the individual activity of NK1R expressing neurons was not affected (Figure 6h–j,m). These findings suggest a role for gap junctions in the maintenance of correlated network activity in the preBötC.

Figure 6 with 1 supplement see all
Gap junctions are necessary to maintain part of the correlated respiratory network.

In the respiratory regions, the gap junction proteins Cx43 (a, N=9), Cx32 (b, c, N=8), and Cx26 (d, arrowheads; double-labeling with NK1R, N=5) are present. Gap junction inhibitors CBX (e) and 18-α-GA (f) reduced network synchronization in the preBötC. Notably, the Ca2+ activity of individual NK1R-positive cells was not affected (hj, m). Correlating cell pair numbers decreased to 21% (N=8) and 20% (N=6) of their respective controls after treatment with CBX and 18-α-GA, respectively (k). Network properties were not affected by GZA, an analog to CBX that lacks the ability to block gap junctions, (g, jk, N=7) or aCSF (N=8). An initial increase in fluorescence intensity was noted after adding CBX and GZA but not after adding 18-α-GA, indicating an immediate excitatory effect of CBX and GZA (l). 18-α-GA reduced the number of active cells in the network at 1 min after application (53%), but CBX did not (91%, N.S.). At the same time point, an increased number of active cells were observed with GZA treatment (139%). After 10 min, a reduction of the number of active cells was found after treatment with both 18-α-GA and CBX (54% and 43%). However, the number of active cells returned to normal after GZA application (89%, N.S.; l). DIV: days in vitro. N: number of slices. Scale bars: 10 µm in a, c, and d, 100 µm in others. Multicolored bar: color-coded correlation coefficient values. Data are presented as means ± SD. *p<0.05. Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.027

Conversely the rhythmic activity of NK1R+ neurons does not depend on gap junctions. Moreover, gap junction inhibition did not affect the mean correlation values, connectivity, or small-world parameter of the remaining correlated cell pairs (Figure 6—figure supplement 1). This demonstrates that the cells connected in a gap junction-independent manner are organized as a small-world network. These results are in line with topological data showing that respiratory neurons are organized in small clusters in the preBötC (Hartelt et al., 2008).

PGE2 modulates preBötC activity

Our in vivo data, as well as others’, indicate that PGE2 and hypercapnia induce sigh activity (Ramirez, 2014; Koch et al., 2015). We hypothesized that this is due to effects on the respiratory centers in the brainstem. We used our brainstem slice cultures of the preBötC to study the direct effects of PGE2 and hypercapnia in vitro.

PGE2 levels in cerebrospinal fluid measured in experimental models and in human infants are in the pico- to nanomolar range (Hofstetter et al., 2007). In the brainstem slice cultures at 7 DIV, the application of PGE2 (10 nM) lowered the Ca2+ signaling frequency of respiratory neurons in the preBötC (Figure 7a–b). PGE2 also induced longer Ca2+ transients, and the signal amplitudes increased compared to those of the controls (Figure 7b). Koch and colleagues (Koch et al., 2015) suggested that the increase in sighs induced by PGE2 is mediated through persistent sodium channels (INaP) (Koch et al., 2015). Indeed, in the preBötC, 10 µM Riluzole, a blocker of the persistent sodium current (INaP), attenuated effect of PGE2 on Ca2+ signal amplitude and length as well as decreasing the signal frequency (Figure 7b). As in previous studies (Toporikova et al., 2015), Riluzole did not affect the Ca2+ signal compared to control periods. Riluzole is used as an INaP blocker, but may also affect other parts of neuronal signaling, such as glutamate release (Wang et al., 2004). Therefore, we cannot completely determine whether the PGE2 effect is due to an effect on the persistent sodium current or interference with glutamate signaling, although an effect on INaP is likely (Koch et al., 2015).

Figure 7 with 1 supplement see all
PGE2 modulates preBötC network activity.

PGE2 lowered the Ca2+ signaling frequency of the preBötC network in WT mice but not in Ptger3-/- mice (ab). The effect was attenuated but not abolished by Riluzole (b). PGE2 also increased signal amplitude and length (ab), an effect that was abolished after Riluzole application (b). Ptger3 is expressed in the preBötC (c, d), and 20% of the EP3Rs were of the α (Gi-protein coupled) subtype and 77% of the γ (Gs-protein coupled) subtype (e). Hypercapnic exposure (pCO2 elevated from 4.6 to 6.6 kPa) did not affect the signal frequency of the preBötC (fg). DIV: days in vitro. Scale bars: 50 µm in c and 10 µm in d. *p<0.05 Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.030

EP3Rs were present in the preBötC (Figure 7c–d). qRT-PCR showed that 20% of the EP3Rs were of the α-subtype (Figure 7e). EP3Rα inhibits adenylate cyclase via Gi-protein, and reduced cAMP levels inhibit FR (Ballanyi et al., 1997). The EP3Rγ subtype, however, which couples to the GS-protein, was the most abundant (Figure 7e).

In vivo, hypercapnia increases sigh activity, VT, FR, and VE (Figure 1). Therefore, we exposed the preBötC brainstem slice culture to increased levels of CO2 by raising the pCO2 levels from 4.6 kPa to hypercapnic 6.6 kPa, while maintaining a constant pH of 7.5 in the aCSF by the addition of bicarbonate. This did not have any effect on the Ca2+ signaling frequency, the Ca2+ signaling pattern or the network structure in wild-type or Ptger3-/- mice (Figure 7f–g, Figure 7—figure supplement 1). However, the preBötC is not the main central chemosensitive region. Instead, the sensitivity to CO2 is more profound in the pFRG. Therefore, we generated organotypic slice cultures of the pFRG/RTN brainstem level.

The pFRG/RTN respiratory region exhibited correlated network activity and retained CO2 sensitivity

The analysis of network structure and function that we conducted on the preBötC was previously not possible to perform in the pFRG/RTN on acute transverse slices. Studies of the pFRG/RTN are particularly interesting because of its crucial role in central respiratory chemosensitivity (Onimaru et al., 2009). We therefore created the same type of brainstem slice culture as with the preBötC slice using slices containing the pFRG/RTN instead (Figure 8a). These brainstem slice cultures expressed neuronal markers as expected (Figure 8b–d, Figure 8—figure supplement 1) and displayed retention of electrical properties, in a manner similar to the preBötC brainstem slice cultures (Figure 8e–f).

Figure 8 with 1 supplement see all
pFRG/RTN brainstem slice culture.

pFRG/RTN slices were selected based on the location of the facial nucleus (VII; a). In the brainstem slice culture, pFRG/RTN expressed the neuronal markers NK1R (b), KCC2 (c), Phox2b (c), vGlut2 (d), and MAP2 (d). The pFRG/RTN neurons also retained adequate electrical properties and generated spontaneous action potentials individually or in clusters (ef). Data are presented as box plots with minimum and maximum values. DIV: days in vitro. Scale bars: 100 µm.

https://doi.org/10.7554/eLife.14170.035

Looking at multiple cells using time-lapse Ca2+ imaging, the activity of the pFRG/RTN was correlated in a scale-free small-world network, akin the one in the preBötC (Figure 9b–d) and was stable during cultivation (Figure 9e–f). There was a slight difference in the number of active cells between 2 week and 3 week cultures (Figure 7e). However, all network properties remained unchanged (Figure 9f and Table 4). The inhibition of neuronal spiking and synapses by TTX (20 nM) disrupted the coordinated activity (21 ± 9% of correlated cell pairs remained, N=11). However, rhythmic Ca2+ activity persisted in a subset of primarily (64 ± 9%, N=11) NK1R-positive cells (Figure 9—figure supplement 1). The pFRG/RTN cells did not exhibit any change in signaling frequency after DAMGO application (Figure 9g, average levels from 7-, 14-, and 21-DIV cultures are displayed, as there were no significant differences among cultures of these ages), confirming the absence of preBötC µ-opioid-sensitive regions in these slices (Ballanyi and Ruangkittisakul, 2009). Similarly to the preBötC brainstem slice culture, the pFRG/RTN responded to higher [K+] with an increase in frequency (Figure 9h; average levels from 7-, 14-, and 21-DIV cultures are displayed, as there were no significant differences among cultures of these ages).

Figure 9 with 2 supplements see all
The pFRG/RTN respiration-related network generates correlated neural activity and responds to CO2.

The pFRG/RTN network is arranged in a small-world manner just ventral to the facial nucleus. The network structure was preserved during cultivation (ad). The number of correlating cell pairs did not change with longer culturing times, but the number of active cells was higher at 3 weeks than at 2 weeks (45 ± 27<76 ± 19, p<0,05; e). The network parameters were stable during cultivation (f). The pFRG/RTN network did not respond to the µ-opioid receptor agonist DAMGO (0.5 µM; n=420, N=4; a), but the average network frequency increased with higher potassium concentrations (22 ± 5 mHz and 38 ± 7 mHz, N=12; b). Both the neural network and individual NK1R/TMR-SP-labeled cells responded to increases in CO2 pressure (pCO2 elevated to 6.6 kPa), indicating that the chemosensitivity was preserved in the pFRG/RTN brainstem slice culture. Suramin, a P2 receptor antagonist, and TNP-APT, a P2X receptor antagonist, attenuated the CO2 response but did not abolish it (g). DIV: days in vitro. Scale bars: 100 µm. Multicolored bar: color-coded correlation coefficient values. N: number of slices, n: number of cells. Data are presented as means ± SD. *p<0.05. Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.038
Table 4

The pFRG/RTN network parameters remain unchanged for 21-DIV cultures. The results of correlation analysis for the pFRG/RTN are shown. Among the analyzed network parameters, only the number of active cells differed at the analyzed time points, and only between 14 and 21 DIV. N.S.: not significant. N: number of slices. Data are presented as mean ± SD.

https://doi.org/10.7554/eLife.14170.045
pFRG/RTN7 DIV (N=12)14 DIV (N=11)21 DIV (N=6)
Correlating cell pairs118 ± 6961 ± 3174 ± 42N.S.
Active cells49 ± 2641 ± 21*76 ± 19**p<0.05
Correlations per active cell3.1 ± 2.21.7 ± 1.11.0 ± 0.7N.S.
Connectivity0.7 ± 0.20.6 ± 0.20.7 ± 0.2N.S.
Mean shortest path length (λ)0.7 ± 0.20.7 ± 0.30.8 ± 0.5N.S.
Clustering coefficient (σ)2.7 ± 1.63.0 ± 1.74.5 ± 3.6N.S.
Small-world parameter (γ)3.6 ± 2.54.2 ± 2.63.3 ± 1.6N.S

Next we examined the CO2 sensitivity of the pFRG/RTN (Onimaru et al., 2008). This resulted in increased signal frequency of the Ca2+ oscillations (Figure 9i, Table 5, Video 2; data from 7-DIV cultures are displayed, and no significant differences in the response among 7-, 14-, and 21-DIV cultures were observed) and the activation of some previously dormant cells. During hypercapnic exposure, the pFRG/RTN network topology remained essentially unchanged (Figure 9—figure supplement 2).

Table 5

pFRG/RTN slices respond to CO2 if the EP3R is present. The average mean frequency of all cells in the network and the average mean frequency of NK1R-positive cells during the control period or during exposure to hypercapnia are shown (pCO2 = 55 mmHg, pH = 7.5). N.S.: not significant. N: number of slices, n: number of cells. Data are presented as mean ± SD.

https://doi.org/10.7554/eLife.14170.046
Mean frequency (mHz)
NetworkNK1R+ cells
ControlHypercapniaControlHypercapnia
pFRG/RTN - WT
(N=7, n=343)
21.6 ± 3.254.1 ± 2.7*p<0.0514.4 ± 0.938.5 ± 5.5*p<0.05
pFRG/RTN - Ptger3−/−
(N=5, n=448)
25.0 ± 7.926.0 ± 1.9N.S.11.4 ± 5.811.6 ± 3.8N.S.
preBötC - WT
(N=5, n=1737)
16.4 ± 2.516.5 ± 1.3N.S.16.6 ± 4.615.7 ± 5.3N.S.
preBötC - Ptger3−/−
(N=4, n=822)
21.1 ± 8.617.3 ± 3.8N.S.22.7 ± 5.917.8 ± 7.7N.S.
Video 2
Ca2+ oscillations visualized with Fluo-4 in the chemosensitive region pFRG/RTN.

Low network activity is increased by exposure to hypercapnia after 15 s. fps: frames per second.

https://doi.org/10.7554/eLife.14170.047

Response to hypercapnia involves pFRG/RTN astrocytes, which release ATP that acts on purinergic P2-receptors (Erlichman et al., 2010; Gourine et al., 2010; Huckstepp et al., 2010a). We sought to examine whether this kind of signaling pathway was active in the 7-DIV brainstem slice cultures, and we found that blocking purinergic receptors with Suramin or TNP-ATP application did not abolish the hypercapnic response, in agreement with previous data (Sobrinho et al., 2014). However, both the unspecific P2 receptor and the more specific P2X receptor antagonist attenuated the CO2 response by approximately one third (30 ± 6%; Figure 9i), as observed in adult and neonatal rats (Wenker et al., 2012) and 9-day-old mice (Gourine et al., 2010). Thus, the CO2-induced release of ATP acting on P2 receptors may contribute to the CO2 response.

In conclusion, our brainstem organotypic slice culture contains an active pFRG/RTN network that retains its structural integrity over time and responds to CO2 exposure with increased activity.

The CO2 response is dependent on EP3R signaling and gap junctions

Gap junctions, both intercellular and hemichannels, are linked to respiratory chemosensitivity (Huckstepp et al., 2010a; Meigh et al., 2013; Reyes et al., 2014). Recently, CO2 was shown to interact with the hemichannel Cx26, inducing an open state through the formation of carbamate bridges, thus increasing the release of compounds such as ATP (Meigh et al., 2013). Therefore, we hypothesized that gap junctions exert functions within the pFRG/RTN network. However, gap junction inhibitors did not affect signaling frequency or network topology of the pFRG/RTN (Figure 10a, Figure 10—figure supplement 1). Instead, the frequency response to hypercapnia was both inhibited and reversed by the application of the gap junction inhibitor 18-α-GA (Figure 10b–c). GZA (a structural analog of CBX without gap junction-inhibiting properties) did not alter the CO2 response (Figure 10b–c).

Figure 10 with 1 supplement see all
Correlated pFRG/RTN network activity is not dependent on gap junctions, but hypercapnic responses are.

Blocking gap junctions in the pFRG/RTN did not change the functional network structure of the respiratory center or alter its frequency (a and c, N=7). However, hypercapnic responses (CO2↑) were abolished when gap junctions were inhibited by 18-α-GA (b, top trace; c, left graph, N=7). GZA (a structural analog of CBX without gap junction-inhibiting properties) increased the frequency, and hypercapnia increased it further (b, middle trace; c, middle graph, N=7). An initiated hypercapnic response was attenuated but not completely reversed by 18-α-GA (b, bottom trace; c, lower graph, N=5). This dynamic was not seen after application of GZA. DIV: days in vitro. Scale bars: 200 µm. N: number of slices. Multicolored bar: color-coded correlation coefficient values. Data are presented as means ± SD. *p<0.05. Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.048

We conclude that 18-α-GA inhibits the hypercapnic response, while inhibition of purinergic signaling pathways attenuates it. Thus, we suggest that the CO2 response is not entirely explained by the connexin-mediated release of ATP. Furthermore, inflammation via PGE2 and EP3R alters the hypercapnic response in vivo and in brainstem spinal cord en bloc preparations (Figure 1 and Siljehav and colleagues Figures 1 and 4 [Siljehav et al., 2014]). Therefore, we hypothesized that hypercapnic responses involve PGE2 signaling and next analyzed the PGE2 content of the aCSF under control and hypercapnic conditions. In all examined slices (N=12/12, 7 DIV), a transient doubling of the PGE2 concentration after pCO2 elevation was evident (Figure 11). When gap junction blockers were applied, this peak was absent (N=4/4, 7 DIV; Figure 11). This indicates a hypercapnia-induced, gap junction-mediated release of PGE2.

Figure 11 with 1 supplement see all
PGE2 is released during hypercapnia.

The aCSF contents exhibited an increase in microenvironmental PGE2 levels during hypercapnia in 12 out of 12 slices. Here, the PGE2 concentration of a brainstem slice culture is displayed during control and hypercapnic periods (a). When gap junctions were inhibited (18-α-GA, blue line), the PGE2 levels remained unaltered during hypercapnia (N=4). The average PGE2 level throughout the whole experiment was not affected by hypercapnia, but the peak value was higher during hypercapnia than under control conditions (b). N: number of slices. Data are presented as means ± SD. *p<0.05. Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.052

Immunohistochemistry showed expression of microsomal prostaglandin E synthase 1 (mPGEs-1) in GFAP positive astrocytes (Figure 11—figure supplement 1). mPGEs-1, the main PGE2 producing enzyme, has previously been found mainly in endothelial cells of the blood brain barrier of adult rats (Yamagata et al., 2001). Our findings suggest that astrocytes in the vicinity of the ventral brainstem border of neonates express mPGEs-1 and might therefore be candidates for modulation of breathing through CO2-induced release of PGE2.

PGE2 has a primarily inhibitory effect on respiration in neonatal mice and humans (Hofstetter et al., 2007), which we confirmed to account for its effects on the preBötC (Figure 7). However, as hypercapnia seems to induce a release of PGE2 while stimulating breathing activity, we hypothesized that PGE2 has a direct stimulatory effect on the pFRG/RTN. Indeed, PGE2 increased the signaling frequency of pFRG/RTN neurons (Figure 12a–b, Table 6). This effect was EP3R dependent, and EP3Rs were present in the pFRG/RTN, expressed both on respiratory neurons and on astrocytes (Figure 12c–e). We also observed a non-significant increase in amplitude (8 ± 3% and 11 ± 4% increase compared to control period, N.S.). Neither the PGE2 effect nor the hypercapnic response of the pFRG/RTN was affected by Riluzole (30 ± 5 mHz vs 25 ± 2 mHz, N.S., N=6, and 36 ± 2 mHz vs 35 ± 6 mHz, N.S., N=6). qRT-PCR showed abundant expression of the EP3Rγ subtype, which couples to the GS-protein (Namba et al., 1993). This would lead to an increase in intracellular cAMP in the pFRG/RTN Ptger3-expressing cells in response to PGE2 (Figure 12f).

PGE2 alters respiratory network activity.

In the pFRG/RTN, PGE2 increased the frequency of respiratory (NK1R-expressing) neurons. This PGE2 effect was absent in brainstem slice cultures lacking EP3R (Ptger3-/-; ab). EP3Rs were present in NK1R-expressing neurons in the pFRG/RTN (c, arrowheads, f) and co-localized with Phox2b (d, arrowheads). EP3Rs were also found on S100B-expressing astrocytes (e, arrowheads). Staining was performed on acutely fixed tissue (ce) and brainstem slice cultures (f). qRT-PCR showed an abundance of the EP3Rγ (Gs-protein coupled) in the pFRG/RTN (N=7; f). N: number of slices. DIV: days in vitro. Scale bars: 100 µm. Data are presented as means ± SD. *p<0.05 Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.055
Table 6

PGE2 increases the frequency of pFRG/RTN neurons and decreases the frequency of preBötC neurons.The mean frequencies of NK1R-positive cells during the control period or during exposure to 10 nM PGE2 are shown. N.S.: not significant. N: number of slices, n: number of cells. Data are presented as mean ± SD.

https://doi.org/10.7554/eLife.14170.057
Mean frequency (mHz)
ControlPGE2
pFRG/RTN - WT
(N=5, n=343)
13.7 ± 1.121.5 ± 2.9*p<0.05
pFRG/RTN - Ptger3−/−
(N=4, n=448)
12.1 ± 2.08.5 ± 2.9N.S.
preBötC - WT
(N=7, n=1737)
20.3 ± 2.28.7 ± 1.4*p<0.05
preBötC - Ptger3−/−
(N=5, n=822)
22.8 ± 2.316.4 ± 1.1N.S.

To further characterize the PGE2 signaling during hypercapnia, we blocked its main receptor, EP3R. Notably, pharmacological blocking of EP receptors (using AH6809, 10 µM) abolished the hypercapnic response (Figure 13a–b, 7 DIV), in line with our in vivo data from Ptger3-/- mice.

pFRG/RTN slices (7 DIV) from Ptger3-/- mice did not respond to hypercapnia (Figure 13c–d). Thus, EP3R is important for pFRG/RTN CO2 responsiveness. We next generated a lentiviral vector in which the mouse EP3R (Ptger3) promoter controls the expression of the red light-activated halorhodopsin Halo57 fused to eGFP (Figure 13e). After transduction, we detected eGFP expression in 90 ± 6% of Phox2b-positive neurons in the pFRG/RTN (Figure 13—figure supplement 1). Stimulation by red (625 nm) light of the transduced brainstem slice cultures (7 DIV) triggered hyperpolarization of Ptger3-halo57-expressing cells and immediately reduced the calcium signaling frequency of both the network and individual NK1R+ neurons (Figure 13—figure supplement 1). This finding indicates a fundamental role for Ptger3-expressing cells in the network. Additionally, the response to hypercapnia in the pFRG/RTN was abolished during the light-induced silencing of Ptger3-expressing cells. The CO2 response was also reversed by the light-induced halo57 hyperpolarization of Ptger- expressing cells (Figure 13f–g, Table 7).

Figure 13 with 1 supplement see all
PGE2, acting through EP3R, is crucial for the hypercapnic response.

Pharmacological inhibition of EP3R by the EP receptor antagonist AH6809 inhibited the response to hypercapnia (increased pCO2[CO2↑]) in the pFRG/RTN (N=6, n=472, N.S.; ab). The hypercapnic response was also absent in pFRG/RTN slices lacking EP3R (Ptger3−/−; N=5, n=348, N.S.; cd). Layout of the lentivirus containing Halo57 (ER2) and eGFP genes under the control of the EP3R promoter (Ptger3) used for optogenetics (WPRE=gene enhancing element; e). During optogenetic silencing of Ptger3-expressing cells, no frequency changes were observed in response to hypercapnia (f, top trace; g, left graph). The hypercapnic response was also reversed by activating Ptger3-Halo57 (f, middle and bottom trace; g, middle and right graph). Red line: Halo57 activation in response to 625 nm light. N: slices, n: cells. Data are presented as means ± SD. *p<0.05. Source data are available in a separate source data file.

https://doi.org/10.7554/eLife.14170.058
Table 7

Silencing of Ptger3-expressing cells inhibits the response to hypercapnia.Mean frequencies of the pFRG/RTN network during the control period and during exposure to hypercapnia with and without Halo57 stimulation are shown. N.S.: not significant. N: number of slices. Data are presented as mean ± SD.

https://doi.org/10.7554/eLife.14170.063
N=41Mean frequency (mHz)
ControlHypercapnia
Control22.9 ± 9.0*34.0 ± 4.3*p<0.05
Halo57 stimulation9.0 ± 1.710.3 ± 1.1N.S.

Based on these findings, we suggest that the PGE2-EP3R pathway is an important mechanism in the hypercapnic response and a modulator of respiratory activity.

Discussion

Here, we present two novel breathing brainstem organotypic cultures in which the respiration-related preBötC and pFRG/RTN regions maintain their functional organization, activity, and responsiveness to environmental cues. Using these cultures, we show that PGE2 is involved in the control of sigh activity and the response to hypercapnia via EP3R in the preBötC and the pFRG/RTN, respectively. These findings provide novel insights into central respiratory central pattern generation, its modulation, and the mechanisms underlying breathing disorders during the neonatal period.

Due to the complexity of the respiratory mechanisms, it is difficult to create optimal in vitro model systems that represent in vivo conditions while allowing sufficient depth in detailed mechanisms and their manipulation. The majority of previous studies were performed on brainstem-spinal cord preparations (en bloc) (Onimaru, 1995) or acute slices (Ruangkittisakul et al., 2006). However, these preparations remain active only for hours, making it difficult to study development and long-term effects on respiratory rhythm. Organotypic slice cultures provide a bridge between cell cultures and animals in vivo (Yamada and Cukierman, 2007). Their preserved three dimensional structure allows functional circuits to be studied and manipulated over time under microenvironmental control (Gähwiler et al., 1997; Gogolla et al., 2006; Yamada and Cukierman, 2007; Preynat-Seauve et al., 2009). First used with hippocampal tissue (Gähwiler, 1988), the organotypic culturing method has since expanded to research on the cerebellum (Lu et al., 2011) as well as on the brainstem auditory circuits (Thonabulsombat et al., 2007). Recently, Phillips and colleagues (Phillips et al., 2016) presented an organotypic model system of the preBötzinger complex with respiration-related neuronal rhythm that persists for a month. Here, we characterize this new type of brainstem slice culture further, and also provide details on respiratory network structure and functional respiratory-related motor output. In addition we show that also the pFRG/RTN retains respiration-related rhythmic activity and chemosensitivity. As with all model systems, it has its limitations, e.g., the slices lose several respiratory-related regions (Smith et al., 2009). Nonetheless, in contrast to acute slices and the brainstem-spinal cord preparation, our new experimental model system allows long-term studies and manipulation of respiratory networks. This enables the use of different techniques and methods, and significantly reduces the number of procedures that otherwise need to be performed on live animals, as well as the total number of experimental animals. We have exploited this advantage by transfecting the brainstem slice cultures in vitro to be suitable for optogenetic techniques.

Using a newly developed cross-correlation analysis algorithm (Smedler et al., 2014), we revealed in the brainstem slice culture, a clustering of cells within the two central pattern generators, a small-world network. A small-world network is characterized by a mean clustering coefficient exceeding that in random networks, but has a mean shortest path-length as short as that in random networks (Watts and Strogatz, 1998; Malmersjo et al., 2013). Furthermore, the presence of the connective nodes and hubs gives the network a scale-free organization. This finding is in line with a previous topological analysis based on neuronal staining in the preBötC (Hartelt et al., 2008). The present insights into the network structure of the pFRG/RTN have not been achieved previously with other methods. Notably, scale-free and small-world networks have been suggested to have evolutionary advantages (Barabasi and Oltvai, 2004; Malmersjo et al., 2013).

Subsequently we examined how the networks and individual cells were connected. Early in development, gap junctions connect the respiration-related fetal neural networks (Thoby-Brisson et al., 2009). During development, gap junction-mediated Ca2+-transients stimulate the proliferation of neural progenitor cells (Malmersjo et al., 2013) and form a template for chemical synapses to coordinate more mature neural networks (Jaderstad et al., 2010). Using CBX and 18-α-GA, we demonstrated that intercellular connections still play a role in postnatal preBötC network activity. This is in line with previous findings (Elsen et al., 2008). Notably, even though fewer cells remained active, respiratory neuron frequency and network structure were not affected. Although both CBX and 18-α-GA are commonly used as gap junction inhibitors (Solomon et al., 2003; Elsen et al., 2008; Véliz et al., 2008; Jaderstad et al., 2010), these drugs have side effects (Rekling et al., 2000; Schnell et al., 2012). We used GZA as a control substance because it is structurally similar to CBX but does not have any gap junction inhibiting properties (Solomon et al., 2003; Li and Duffin, 2004; Elsen et al., 2008). However, it mimics many of the side effects of CBX, e.g. the initial stimulatory effect seen in the present study. These limitations need to be kept in mind when interpreting our results on gap junction functions, and further studies are needed to confirm them, preferably using more specific methods of connexin blockage, such as RNAi.

However, our findings do suggest the presence of a neuron-specific subnetwork, connected by chemical synapses, that is able to maintain the network structure. Furthermore, another subnetwork, likely a glial one (Giaume et al., 2010; Okada et al., 2012) driven by the electrical connections that modulate network output also seems to be present. Thus, neonatal preBötC synchronization is both gap junction-and synaptic signal-dependent (Feldman and Kam, 2015), and it probably contains both neuronal and glial subnetworks. The pFRG/RTN, by contrast, requires gap junctions for its establishment in rodents but is not dependent on them postnatally for rhythmic, correlated network activity (Fortin and Thoby-Brisson, 2009). The main mechanism that drives activity in the pFRG/RTN is glutamatergic (Guyenet et al., 2013). By contrast, pFRG/RTN gap junctions seem here to be involved in the hypercapnic response (Figure 10 and 11). It has been suggested that Cx26 is directly modulated by CO2, independent of H+, through the formation of carbamate bridges (Meigh et al., 2013). Our data do not distinguish between intracellular pH-dependent and -independent mechanisms. However, since PGE2 can pass through connexins (Reyes et al., 2014), the present data are in line with a CO2-induced, connexin-mediated, release of PGE2 (Figure 14).

Model of how PGE2 modulates respiration and sighs in the preBötC and pFRG/RTN.

Systemic inflammation, through the proinflammatory cytokine IL-1β and hypoxia, induces the production of PGE2 in blood brain barrier (BBB) endothelial cells (Hofstetter et al., 2007).PGE2 subsequently induces respiratory depression and increases sigh activity via the inhibitory G-protein coupled receptor EP3Rα in the preBötC. In the pFRG/RTN, PGE2 plays a role in the response to elevated pCO2. CO2 directly modulates connexin 26 (Cx26) hemichannels, leading to ATP release. The results in this study suggest that Cx26 also releases PGE2, possiblyfrom mPGEs-1+ astrocytes. PGE2 increases respiratory activity via the stimulatory G-protein coupled receptor EP3Rγ on pFRG/RTN neurons. Thus, inflammation, hypoxia, and hypercapnia alter respiratory neural network and motor output and breathing activity through distinct effects of PGE2 in the pFRG/RTN and the preBötC, respectively. Chronically elevated PGE2 levels, as observed during ongoing inflammation, may decrease the central pattern generators’ ability to respond to hypoxic and hypercapnic events. In extreme cases, this decrease may have fatal consequences.

https://doi.org/10.7554/eLife.14170.064

Prostaglandins are important regulators of autonomic functions in mammals. In many disease states, acute inflammatory responses are initially protective but become harmful under chronic conditions. In our previous reports, we demonstrated how the pro-inflammatory cytokine interleukin (IL)-1β impairs respiration during infection by inducing a PGE2 release in the vicinity of respiratory centers. We also showed that infection is the main cause of respiratory disorders in preterm infants (Hofstetter et al., 2007, 2008) and, in the case of apneas, bradycardias and desaturations (ABD) events in neonates (Siljehav et al., 2015). PGE2 is also a key component in the regulation of sigh frequency (Ramirez, 2014; Koch et al., 2015). During and immediately after birth, PGE2 levels are increased (Mitchell et al., 1978). Indeed, the first breaths of extrauterine life are deep and sigh-like, facilitating alveolar recruitment and CO2 removal (Mian et al., 2015). In the brainstem slice cultures, PGE2 had a direct EP3R-dependent effect on both respiratory centers. Notably, PGE2 increased pFRG/RTN but inhibited preBötC frequency (Video 3). This finding might be explained by the different distributions of EP3R subtypes in the different regions (Figure 12). The coupling to inhibitory or stimulatory G proteins depends on the alternative post-transcriptional splicing of the C-terminal tail of the EP3R preprotein (Namba et al., 1993). Furthermore, PGE2 caused a longer Ca2+ transient and a higher relative amplitude in an INaP-dependent manner, mimicking the PGE2-based induction of sighs that we observe in vivo and that were recently reported by Koch and colleagues in acute preBötC slices (Koch et al., 2015).

Video 3
Parallel display of Ca2+ oscillations visualized with Fluo-4 in the pFRG/RTN (left) and preBötC (right).

After 15 s, PGE2 (10 nM) is added. This increases the activity of the pFRG/RTN network while the preBötC activity is inhibited. fps: frames per second. For high-resolution versions of the videos, please follow this link to the Karolinska Institutet Cloud Storage system (Box): https://ki.box.com/s/abzuei0yzl4dzbn99995382va6btsq4l.

https://doi.org/10.7554/eLife.14170.065

Recent data reveal a role of neuromedin B (NMB) and gastrin-related peptide (Grp) and NMB-GPR-expressing preBötC neurons in sighing (Li et al., 2016). In addition to these peptidergic pathways, the present and recent data from Koch and colleagues (Koch et al., 2015) suggest that low concentrations of the inflammation-associated PGE2 induce sighs, acting through modulation of the persistent sodium current in preBötC neurons.

The preBötC results presented in this study provide evidence for how the general respiratory depression induced by inflammatory signaling, previously reported in vivo and in vitro(Hofstetter et al., 2007) and in human neonates (Hofstetter et al., 2007; Siljehav et al., 2015), is mediated by a direct effect of PGE2 on EP3R (Siljehav et al., 2012) in the preBötC. The present data may help to further explain the mechanism underlying apneas that occur during infectious periods in neonates (Hofstetter et al., 2007, 2008; Di Fiore et al., 2013; Siljehav et al., 2015).

Another common respiratory problem in neonates, particularly premature infants, is an inability to respond adequately to hypoxia and hypercapnia. This may cause recurrent hypoxia, leading to cognitive disabilities later in life (Greene et al., 2014). A disruption of central CO2 chemosensitivity is commonly seen in children with bronchopulmonary dysplasia (Di Fiore et al., 2013), leading to chronic hypoventilation, which may explain why these infants have an increased risk of sudden infant death syndrome (Martin et al., 2011). Therefore, we investigated the role of the pFRG/RTN in chemosensitivity (Guyenet et al., 2013) and found that the response to hypercapnia is dependent on functioning gap junctions. This is in line with previous findings showing that Cx26 is directly modified by CO2 (Meigh et al., 2013).

These CO2-sensitive connexin hemichannels can release ATP, and indeed the hypercapnic response is partly mediated by purinergic type 2 receptors (Erlichman et al., 2010; Gourine et al., 2010; Guyenet et al., 2013). In addition to these purinergic pathways, we suggest that EP3R-dependent signaling is involved in the response to altered pCO2. Genetic ablation of Ptger3 reduced the hypercapnic response both in vivo and in vitro, as did pharmacological blockage in vitro, in line with our previous experiments (Siljehav et al., 2014). Moreover, the optogenetic inhibition of Ptger3-expressing cells in the pFRG/RTN revealed that these cells are essential for the CO2 response. We also demonstrated that PGE2 is released during hypercapnic exposure, likely through Cx26 or other CO2-sensitive connexins (Huckstepp et al., 2010b). Thus, part of the CO2 response seems to be mediated by a gap junction-dependent release of PGE2.

Generation of active expiration is another important function of the pFRG/RTN (Feldman et al., 2013). It is possible that PGE2 stimulates both chemosensitive neurons and neurons important for active expiration. Such neuronal populations could overlap, but the ventral part pFRG/RTN seems to have a more chemosensitive character while the lateral part displays rhythmic activity and enforces active expiration when stimulated (Pagliardini et al., 2011; Feldman et al., 2013; Huckstepp et al., 2015). The CO2-sensing of the pFRG/RTN slice remains functional. Whether the rhythmic activity we observe in the pFRG/RTN is generated by “active expiration-neurons” is outside the scope of the present study. Future studies should aim to investigate whether PGE2 also may affect active expiration.

The pFRG/RTN is the best-recognized central chemosensitive region. However, in our pFRG/RTN brainstem slice culture, neurons of the raphe nucleus should be present (Smith et al., 2009). Such neurons may also have chemosensing properties (Richerson, 2004), though this has not been shown conclusively (Depuy et al., 2011). From the raphe nucleus there are evidence of projections to the pFRG/RTN (Guyenet et al., 2009), and we cannot exclude the possibility that these are preserved in the brainstem slice culture.

The effects of CO2 in the present study are based on a change in carbamylation of specific proteins, e.g. Cx26 (Meigh et al., 2013), or intracellular pH, but testing these alternatives goes beyond the scope of the present work. In our experimental setup the extracellular pH remained stable while the dissolved CO2 increased. This specific approach was selected because CO2 has a direct modulating effect on connexins, allowing passage of small molecules (Huckstepp et al., 2010a; Huckstepp and Dale, 2011; Meigh et al., 2013), and our hypothesis was that PGE2 is released through such connexins.

What still remains to be determined the exact source of the PGE2 released during hypercapnia. The indication of a gap junction-dependent release of PGE2 together with the presence of mPGEs-1 in pFRG/RTN astrocytes suggests that the PGE2 is of astrocytic origin. This would be in line with previous findings of astrocytic ATP release during hypercapnia (Gourine et al., 2010; Huckstepp et al., 2010a). The astrocytic involvement in the CO2 response is also evident in a Rett syndrome model (methyl-CpG-binding protein 2 (MeCP2) knockout), in which conditional MeCP2 knockout in astroglia blunts the CO2 response (Turovsky et al., 2015). We think that mPGEs-1-expressing astrocytes are the likely source, even though alternative sources of PGE2, such as endothelial cells or microglia, remain to be investigated with regards to their possible involvement in the pFRG CO2 response. Nonetheless, CO2-mediated PGE2 release introduces a novel chemosensitive pathway (Figure 14).

As PGE2 and the EP3R are directly involved in and modulate both the respiratory rhythm-generating preBötC and the Phox2b chemosensitive neurons, PGE2 from other sources, such as endothelial cells during hypoxia and inflammation (Hofstetter et al., 2007), will alter the hypercapnic and the hypoxic responses. PGE2 has prominent respiratory depressant effects in humans, sheep, pigs, and rodents (Guerra et al., 1988; Long, 1988; Ballanyi et al., 1997; Hofstetter et al., 2007; Siljehav et al., 2015). The PGE2-induced attenuation of these vital brainstem neural networks, e.g., during an infectious response, could result in gasping, autoresuscitation failure and ultimately death. However, how chronic PGE2 release associated with ongoing inflammation alters plasticity and the responsiveness to CO2 must be further investigated.

To conclude, we identified a novel pathway in the hypercapnic response of brainstem neural networks that control breathing. This pathway depends on EP3R and gap junctions and is partly mediated by the release of PGE2, linking chemosensitivity control to the inflammatory system. The present findings have important implications for understanding why and how ventilatory responses to hypoxia and hypercapnia are impaired and inhibitory reflexes exaggerated in neonates, particularly during infectious episodes.

Materials and methods

Subjects

C57 black (C57BL/6J) inbred mice (Charles River, Wilmington, MA) were utilized in the experiments. The eicosanoid prostanoid 3 receptor (EP3R) gene (Ptger3) was selectively deleted in knockout mice (Ptger3−/−) with a C57BL/6J background, as described preciously (Fabre et al., 2001). C57BL/6J mice were then used as experimental controls for Ptger3−/− mice. As results from Ptger3−/− mice were consistent with pharmacological and optogenetic inhibition of EP3Rs, we can confirm the lost EP3R function in the mice.

To determine the location of mPGEs-1, mice expressing green fluorescent protein (GFP) under the GFAP promoter were used. Frozen sperm from the GFAP-tTA (Lin et al., 2004; Pascual et al., 2005) and tetO-Mrgpra1 (Fiacco et al., 2007) mouse strains were purchased from the Mutant Mouse Regional Resource Centers supported by NIH (MMRRC). The strains were re-derived by Karolinska Center for Transgene Technologies (KCTT), and the offspring was crossed as previously described (Fiacco et al., 2007). Double transgenics were identified by PCR according to MMRRC's instructions.

All mice were reared by their mothers under standardized conditions with a 12:12-hr light-dark cycle. Food and water was provided ad libitum. The studies were performed in accordance with European Community Guidelines and approved by the regional ethic committee. The animals were reared and kept at the Department of Comparative Medicine, Karolinska Institutet, Stockholm, Sweden.

Dual-chamber plethysmography in vivo

Ventilatory measurements were made using dual-chamber plethysmography in 9-day old (P9) mice. Mice were cooled on ice for 2–3 min and then prostaglandin E2 (PGE2, 1 µM; Sigma-Aldrich, St. Louis, MO, USA, cat no. P5640) or vehicle (artificial cerebrospinal fluid, aCSF, containing in mM: 150.1 Na+, 3 K+, 2 Ca2+, 2 Mg2+, 135 Cl, 1.1 H2PO4, 25 HCO3- and 10 glucose) was slowly injected into the lateral ventricle by using a thin pulled glass pipette attached to polyethylene tubing (Siljehav et al., 2014). The mouse was then immediately placed into the plethysmograph chamber. After a 10-min recovery period, confirming stable respiration and body temperature, respiratory parameters in normocapnia (air) was established followed by a hypercapnic challenge (5% CO2 and 20% O2 in N2) for 5 min. This was followed by 5 min of normocapnia. Skin temperature was measured throughout experimentation and remained stable. After experimentation, the mice were anesthetized with 100% CO2 and decapitated. The brain was dissected and examined at the injection site and for the presence of any intracranial hemorrhage. Three of 28 animals had visible intracranial bleeding and were excluded from analysis.

Brainstem organotypic culture

P3 mice pups were used for the establishment of brainstem organotypic slice cultures. The pups were decapitated at the cervical C3–C4 level. The heads were washed with cold dissection medium consisting of 55% Dulbecco’s modified Eagle’s medium (Invitrogen, Paisley, UK), 0.3% glucose (Sigma-Aldrich, St. Louis, MO, USA), 1% HEPES buffer (Invitrogen, UK) and 1% Antibiotic-Antimycotic (Invitrogen, UK). After washing, the heads were moved to fresh dissection medium on ice. The entire brain was dissected. During dissection, extra caution was taken around the cerebellopontine angle to ensure that the respiratory regions of the brainstem were not damaged. Nerves were cut with microscissors.

The brain was sectioned into 300-µm-thick transverse slices by using a McIlwain Tissue Chopper (Ted Pella, Inc., Redding, CA, USA). Slices were selected by using anatomical landmarks, such as the shape and size of the entire slice and the fourth ventricle. For location of the preBötzinger complex (preBötC), the presence of nucleus hypoglossus, nucleus spinalis nervi trigemini, pyramis medullae oblongatae and nucleus tractus solitarius (not always clearly seen), together with the absence of the anterior horn for the nucleus cochlearis, according to online references (Ruangkittisakul et al., 2006, 2011, 2014). For location of the parafacial respiratory group/retrotrapezoid nucleus (pFRG/RTN), the presence of the nucleus facialis was used. On the slices, the preBötC is located within ventrolateral regions, and the pFRG/RTN is located at the ventrolateral edge.

Selected slices were washed by moving them to brain slice medium (55% Dulbecco’s modified Eagle’s medium, 32.5% Hank’s balanced salt solution, 0.3% glucose, 10% fetal bovine serum, 1% HEPES buffer and 1% Antibiotic-Antimycotic [Invitrogen, UK]), after which they were carefully placed on insert membranes (Millicell Culture Plate Inserts; Millipore, Billerica, MA, USA) in six-well plates. The membranes were coated in advance with poly-L-lysine (0.3 ml; 0.1 mg/ml, Sigma-Aldrich, St. Louis, MO, USA). Brain slice medium (1 ml) was placed underneath the membrane, and all fluid on top of the membrane was removed. It is important not to cover the slices with medium, because this may impair oxygenation (Frantseva et al., 1999). The brainstem slice cultures were maintained in an incubator (37°C, 5% CO2), and the brain slice medium was changed every second day. The brainstem slices were kept in culture for 7–21 days in vitro (DIV) before fixation or live imaging experiments. For a detailed protocol, see Herlenius and colleagues (Herlenius et al., 2012).

Immunohistochemistry

For immunohistochemistry, brainstem slice cultures were fixed with cold paraformaldehyde (4%) in PBS for 1 hr at 4°C and 20% ice-cold methanol in PBS for 10 min. Permeabilization was conducted by using 0.2% Triton X-100 (Roche Diagnostics, Hofgeismar, Germany) and 0.1% Tween 20 (Invitrogen, UK) in PBS for 40 min at room temperature (RT). Thereafter, slices were blocked in 5% bovine serum albumin (BSA; Invitrogen, UK) and 0.05% Tween 20 in PBS for 2 hr at RT. The Millicell insert membranes were carefully cut with a scalpel and placed back into the wells. The primary antibodies were diluted 1:200 in 0.05% Tween 20/PBS and incubated at 4°C for 48 hr. Next, the slices were washed 3 × 10 min with PBS and incubated for 1.5 hr at RT with Alexa Fluor-conjugated secondary antibodies (Invitrogen, UK) diluted 1:200 in 0.05% Tween 20/PBS. The slices were then washed 3 × 10 min with PBS and mounted in ProLong Gold Antifade Reagent with DAPI (Invitrogen, UK, cat. no. P36931). Primary antibodies used were mouse anti-microtubule associated protein 2 (MAP2; Invitrogen, cat. no. P11137), rabbit anti-neurokinin 1 receptor (NK1R; Sigma-Aldrich, St. Louis, MO, USA, cat no. S8305), mouse anti-GFAP (Chemicon, Temecula, CA, USA, cat no. MAB360), rabbit anti-S100β (Millipore; cat. no. 04–1054), mouse anti-neuron-specific class III β-tubulin (Tuj1; Covance, Princeton, NJ, USA, cat no. MMS-435P), rabbit anti-K+/Cl cotransporter 2 (KCC2; Millipore, cat no. 07–432), rabbit anti-vesicular glutamate transporter 2 (VGLUT2; Synaptic Systems, Goettingen, Germany, cat no. 135–402), mouse anti-connexin 26 (Cx26; Invitrogen, Inc., San Francisco, CA, cat no. 13–8100), rabbit anti-connexin 32 (Cx32; Invitrogen, cat. no. 71–0600), mouse anti-connexin 43 (Cx43; Zymed, cat no 13–8300), goat anti-Phox2b (Santa Cruz Biotechnology, Santa Cruz, CA, USA, cat no 13224), goat Phox2b antibody (R & D Systems, Minneapolis, MN, USA), and rabbit anti-caspase 3 (Cell Signaling Technology, Beverly, MA, USA, cat no. 9661). Negative controls with only secondary antibodies showed no staining.

For EP3R staining, a different protocol was used. Initially, brains were fixed with 4% paraformaldehyde overnight followed by 10% sucrose overnight and then frozen to -80%. The frozen brainstems were cryosectioned and blocked in blocking buffer (1% BSA, 5% donkey serum, 5% dimethyl sulfoxide (DMSO), 1% Triton X-100 in Tris-buffered saline (TBS, consisting of 6 mM Tris-HCl, 1 mM Tris base and 9 mM NaCl in ddH2O) for 1 hr at RT. After blocking, the slices were incubated with polyclonal rabbit anti-EP3R antibody (Cayman Chemical Co., Ann Arbor, MI, USA) diluted 1:50 in 10% DMSO containing 0.2% Triton X-100 in TBS at RT overnight. Next, slices were washed 3 × 15 min with TBS with agitation, followed by incubation for 1 hr in the dark with Alexa Fluor 488-conjugated donkey anti-rabbit secondary antibody (Life Technologies, Grand Island, NY, USA) diluted 1:1000 in 1% BSA, 2% donkey serum, 2% DMSO and 5% Triton X-100 in TBS. The slices were then washed 3 × 15 min with TBS with agitation, and blocked again for 1 hr at RT in the same blocking buffer as used previously. After blocking, the slices were incubated with the second primary antibody, diluted 1:200 in 10% DMSO containing 0.2% Triton X-100 in TBS at 4°C overnight. Following overnight incubation, the slices were washed 3 × 15 min with TBS with agitation and incubated with Alexa Flour 647-conjugated donkey anti-goat secondary antibody (Life Technologies, Grand Island, NY, USA) diluted 1:1000 in 1% BSA, 2% donkey serum, 2% DMSO and 5% Triton X-100 in TBS. Finally, the slices were washed 3 × 15 min with TBS with agitation, and mounted in ProLong Gold Antifade Reagent with DAPI.

Antibody binding was controlled by including an irrelevant rabbit polyclonal IgG isotype control (Bioss, Woburn, MA, USA). EP3R staining was controlled by including an EP3R blocking peptide reconstituted in distilled water mixed with EP3R antibody at a 1:1 (v/v) ratio. A pre-incubation of EP3R antibody with the blocking peptide for 1 hr at RT was necessary before the antibody was added to the slice. The peptide was used in conjunction with the antibody to block protein-antibody complex formation during immunohistochemical analysis for the EP3Rs. These controls showed no staining.

Double immunofluorescence staining was also performed according to Westman and colleagues (Westman et al., 2004) using polyclonal rabbit anti-human microsomal prostaglandin E synthase 1 antiserum (mPGES-1; Cayman chemicals, cat. no. 160140) and monoclonal anti-mouse glial fibrillary acidic protein antibody (GFAP; Chemicon, Temecula, CA, USA, cat no. MAB360). PBS supplemented with 0.1% saponin (PBS-saponin) was used as a buffer through the experiment. Endogenous peroxidase activity was blocked using PBS containing 1% H2O2 and 0.1% saponin for 60 min in darkness. Endogenous biotin was blocked using an avidin-biotin blocking kit (Vector Laboratories, Burlingame, CA) supplemented with 0.1% saponin. The sections were incubated with primary antibodies overnight, in PBS-saponin containing 3% BSA antibody solution. Thereafter, they were blocked with 1% normal goat serum, or normal donkey serum (depending on the host of secondary antibody) in PBS-saponin for 15 min, followed by 1-hr incubation with secondary antibody, donkey anti-rabbit alexa fluorophore 488 or goat anti-mouse Alexa Fluor 546.

Propidium iodide staining

Propidium iodide (1 ml/L, Invitrogen, UK) was added to brain slice medium (dilution 1:1000). Staining solution (1 ml) was added on top of the membrane with the brainstem slice cultures and incubated at 37°C (5% CO2) for 3 hr. Immediately after incubation, the brainstem slice cultures were fixed in 4% paraformaldehyde for 1 hr. Positive controls were made by first treating the brainstem slice culture for oxygen glucose deprivation (OGD) for 1 h, as described by Montero Dominguez and colleagues (Montero Domínguez et al., 2009).

Electrophysiology

Whole-cell patch recordings were obtained from brainstem slice cultures at a temperature of 34°C. Cells were visualized by using IR-differential contrast microscopy (Axioskop FS, Carl Zeiss, Jena, Germany). Recorded cells were selected visually, and paired recordings were obtained for neurons with lateral somatic distances of <100 µm. Recordings were amplified by using 700B amplifiers (Molecular Devices, Sunnyvale, CA, USA), filtered at 2 kHz, digitized at 5–20 kHz by using ITC-18 (Instrutech, Longmont, CO, USA), and acquired by using Igor Pro (Wavemetrics, Lake Oswego, OR, USA). Patch pipettes were pulled with a P-97 Flamming/Brown micropipette puller (Sutter Instruments, Novato, CA, USA) and had an initial resistance of 5–10 MΩ in a solution containing in mM: 110 K-gluconate, 10 KCl, 10 HEPES, 4 Mg-ATP, 0.3 GTP and 10 phosphocreatine. Recordings were performed in current-clamp mode, with access resistance compensated throughout the experiments. Recordings were discarded when access resistance increased beyond 35 MΩ. To characterize the electrical properties of the recorded cells, depolarizing and hyperpolarizing current steps and ramps were injected, enabling the extraction of properties such as input resistance, membrane time constant and action potential threshold. Electrophysiological properties were presented as box plots, with maximum and minimum values.

For recording of hypoglossal nerve activity and hypoglossal nucleus neuronal population discharge, an extracellular suction electrode was used together with a Model 1700 AC amplifier (A-M systems, Carlsborg, WA, USA) and AxoScope software, version 9.2 (Axon Instruments, Union City, CA, USA). Recordings were made with a sampling interval of 0.3 ms.

Ca2+ time-lapse imaging

For Ca2+ imaging, Fluo-4 AM (Invitrogen, UK) dissolved in DMSO (Invitrogen, UK) was used at 10 µM in serum free brain slice medium or artificial cerebrospinal fluid (aCSF, containing in mM: 150.1 Na+, 3 K+, 2 Ca2+, 2 Mg2+, 135 Cl, 1.1 H2PO4, 25 HCO3- and 10 glucose) together with 0.02% pluronic acid (Invitrogen, UK). We did not observe any differences in Ca2+activity between HEPES-free brain slice medium and aCSF during Ca2+imaging, despite slight differences in [K+] and [Ca2+], which both affect the rhythm of the slice (Ballanyi and Ruangkittisakul, 2009). A higher [Ca2+] or an increase in [K+] from 3 mM to 4.8 mM did not affect the network properties in our system. To localize the preBötC or the pFRG/RTN, tetramethylrhodamine-conjugated Substance P (TMR-SP; Biomol, Oakdale, NY, USA) was used at a final concentration of 3 µM in brain slice medium or aCSF. The TMR-SP solution was placed on top on the brainstem slice and incubated for 10–12 min at 37°C in an atmosphere of 5% CO2. The TMR-SP solution was then replaced with 1 ml of 10 µM Fluo-4 solution. The Fluo-4 solution was incubated for 30–40 min (37°C, 5% CO2). Before imaging, the slice was washed with brain slice medium/aCSF for 10 min (37°C, 5% CO2).

During time-lapse imaging, slices were kept in an open chamber perfused with HEPES-free brain slice medium (containing in mM: 132 Na+, 4.8 K+, 1.4 Ca2+, 0.74 Mg2+, 112 Cl-, 0.76 H2PO4-, 25.6 HCO3- and 16.8 glucose) or aCSF (2.5 ml/min) by using a peristaltic pump. A Chamlide Inline Heater (Live Cell instruments, Seoul, Korea, cat no. IL-H-10) was used for temperature control, and a Chamlide AC-PU perfusion chamber for 25-mm coverslips (Live Cell instruments, Seoul, Korea, cat no. ACPU25) was used for perfusion. HEPES-free medium was used to minimize the risk for hydrogen peroxide formation (Lepe-Zuniga and Gery, 1987). The medium or aCSF was constantly bubbled with 5% CO2 and 95% O2. The temperature of the chamber was set to 32°C, which Hartelt and colleagues (Hartelt et al., 2008) showed to be well tolerated by neurons. Images were captured by using a Zeiss AxioExaminer D1 microscope equipped with 20× and 40× water immersion objectives (N.A. 1.0), a monochromatic Zeiss MrM CCD-camera, a Photometrics eVolve EMCCD-camera and filter sets 38HE (Zeiss), 43 (Zeiss), and et560/hq605 (Chroma, Bellows Falls, VT, USA). For live imaging, a frame interval of 0.1–2 s was used. Exposure time was set to 100–300 ms.

Substances added during imaging were [D-Ala2, N-Me-Phe4, Gly5-ol]-enkephalin (DAMGO, 0.5 µM; Sigma-Aldrich, St. Louis, MO, USA, cat no. E7384), carbenoxolone (CBX 50, µM; Sigma-Aldrich, St. Louis, MO, USA, cat no. C4790), 18α-glycyrrhetinic acid (18-α-GA, 25 µM; Sigma-Aldrich, St. Louis, MO, USA, cat no. G10105), glycyrrhizic acid (GZA, 50 µM; Sigma-Aldrich, St. Louis, MO, USA, cat no. 50531), tetrodotoxin (TTX, 20 nM, Abcam, Cambridge, UK, cat.no. 120055), riluzole (10 µM, Sigma-Aldrich, St. Louis, MO, USA, cat.no. R116), flufenamic acid (FFA, 50 µM, Sigma-Aldrich, St. Louis, MO, USA, cat.no. F9005), Suramin (100 µM; Sigma-Aldrich, St. Louis, MO, USA, cat no. S2671), TNP-ATP (20 nM; Sigma-Aldrich, St. Louis, MO, USA, cat. no. SML0740), AH6809 (Cayman Chemicals, cat.no. 33458-93-4) and prostaglandin E2 (PGE2, 10 nM; Sigma-Aldrich, St. Louis, MO, USA, cat no. P5640). All substances were dissolved in brain slice medium/aCSF prior to experimentation and added to the chamber by using a continuous flow system. For each experiment, a control period with regular medium/aCSF was followed by drug application. GZA was used as a negative control for the gap junction inhibitors CBX and 18α-GA because it has non-gap junction-inhibiting properties, but similar side effects to those of CBX. Specificity was tested by using a second batch of medium or aCSF. During infections in neonatal children, PGE2 is present at a concentration of 15 pM in cerebrospinal fluid (Hofstetter et al., 2007). A higher concentration (10 nM) was used to compensate for the in vivo metabolism of the molecule.

Exposure to isohydric hypercapnia was done by using aCSF adjusted with a high bicarbonate buffer concentration (in mM: 150.1 Na+, 3 K+, 2 Ca2+, 2 Mg2+, 111 Cl, 1.1 H2PO4, 50 HCO3 and 10 glucose). This generated a hypercapnic carbon dioxide partial pressure (pCO2) of 6.6 kPa at pH 7.5 when aCSF was saturated with 8% CO2.

Viral transfection and optogenetics

A subgroup of 1-DIV-old brainstem slices were moved to a separate BSL-2 laboratory where they were transduced with a mouse prostaglandin E receptor 3 (subtype EP3) lentivirus (Ptger3) containing Halo57, developed in collaboration with Dr Robert Finney (Xactagen, Shoreline, WA, USA), by applying 0.2 µl of virus suspension on top of the slice. The brainstem slice cultures were then placed in an incubator for 5 days, and after washing with warm brain slice medium at time points 2 and 5 days, the brainstem slice cultures were moved back to the original laboratory and placed in an incubator overnight. Ca2+ time-lapse imaging was performed on the slices as described above. Halo57 was stimulated continuously during Ca2+ time-lapse imaging by using a 625-nm LED in a custom-built system (Thorlabs, Newton, NJ, USA).

The optogenetically inhibited network and NK1R positive neurons retained their response to general depolarization induced by elevated [K+] (Supplementary Fig. S6).

PGE2 ELISA

The release of PGE2 in aCSF during control and hypercapnic conditions was assessed by ELISA. The aCSF samples were collected through perfusion system, during control and hypercapnic period and either analyzed immediately or stored at -80°C. For the validation of the experiments, two different ELISA kits have been used.

Prostaglandin E2 EIA monoclonal kit by Cayman Chemical (Ann Arbor, MI, US) was performed according to standard procedure. Firstly, the PGE2 EIA Standard was prepared from #1 to #8. The 96-well plate was ready to use and contained a minimum of two blanks (Blk), two non-specific binding wells (NSB), two maximum binding wells (B0) and an eight point standard curve run in duplicate. Each sample was assayed in triplicate. The 96-well plate was coated for 18 hr at 4°C with 50 μl of Prostaglandin E2 AChE Tracer and 50 μl of Prostaglandin E2 Monoclonal Antibody per well. Plate was washed three times with specific Wash Buffer and in consequence, it was developed in the dark at room temperature on a plate shaker for 60–90 min by adding 200 μl of Ellman’s Reagent to each well. Finally, the plate was read at 405 nm.

PGE2 ELISA kit by Enzo Life Sciences (Farmdale, NY, US) was also used for the confirmation of the results. A similar process was followed but a bit shorter. Samples were assayed in duplicate. The 96-well plate was incubated at room temperature on a plate shaker for 2 hr with 50 μl of PGE2 conjugate and 50 μl of antibody solution per well. Then, the plate was washed three times with washing solution. After the wash, 200 μl of the pNpp substrate solution were added to every well and the plate was incubated at room temperature for 45 min. Finally, 50 μl of Stop Solution were added to every well in order to stop the reaction and the plate was read immediately at 405 nm.

Quantitative real-time PCR

The preBötC and pFRG/RTN regions were cut out from brainstem slices with micro scissors. The samples were pooled together litterwise to minimize the effect of different tissue piece sizes, and provide enough cells for accurate analysis. RNA was isolated from the tissue samples using the miRCURY RNA isolation Kit (Exiqon) according to manufacturer’s instructions. cDNA was synthesized from 20 ng RNA using SuperScript VILO cDNA Synthesis Kit (Invitrogen). The reverse transcription was performed according to the manufacturer’s protocol. Real-time PCR was run with Power SYBR Green PCR Master Mix (Applied Biosystems) and amplified in a 7500 Real Time PCR system (Applied Biosystems). Primers are listed in Table 8. As endogenous control, glucose-3 phosphate dehydrogenase (GAPDH; Applied Biosystems) was used. Relative quantification (RQ) values were calculated using the CT(ΔΔCT) method (Livak and Schmittgen, 2001).

Table 8

Primers used for qRT-PCR.

https://doi.org/10.7554/eLife.14170.066
Oligo nameSequence
EP3alfa forward
EP3alfa reverse
GCTTCCAGCTCCACCTCCTT
CATCATCTTTCCAGCTGGTCACT
EP3 sense
EP3beta anti-sense
5′-TGACCTTTGCCTGCAACCTG-3′
5′-GACCCAGGGAAACAGGTACT-3′
EP3gamma forward
EP3gamma reverse
AGTTCTGCCAGGTAGCAAACG
GCCTGCCCTTTCTGTCCAT

Data analysis

From in vivo plethysmograph recording (LabChart Pro, v 8.0.10, AD Instruments, Dunedin, New Zealand), periods of calm respiration without movement artifacts were selected for analysis based upon visual observations during experimentation as in previous studies (Hofstetter and Herlenius, 2005). Mean respiratory frequency (FR; breaths/min), tidal volume (VT) and minute ventilation (VE) during normocapnic and hypercapnic periods were calculated as described previously (Hofstetter and Herlenius, 2005). Sighs were excluded from the analysis. VT and VE were divided by body weight (BW) and expressed as milliliters per gram and milliliters per gram per min, respectively. The number of sighs, defined as breath with larger amplitude and a biphasic inspiratory phase, was calculated manually and expressed as sighs per min.

Immunohistochemical staining was analyzed in a Zeiss AxioExaminer D1 microscope (10×, 20× and 40× water immersion objectives) or a Zeiss LSM700 confocal (40× and 63× oil-immersion objectives), and captured images were processed by adjusting contrast in ImageJ (1.42q, National Institutes of Health, Bethesda, MD, USA) to reduce background staining.

Ca2+ imaging time traces were analyzed with a recently published method (Malmersjo et al., 2013; Smedler et al., 2014). Regions of interest were marked for all cells based on the standard deviation of fluorescence intensity over time, by using a semiautomatic-adapted ImageJ script kindly provided by Dr. John Hayes (The College of William and Mary, Williamsburg, VA, USA, http://physimage.sourceforge.net/). The mean intensity value and coordinates were measured using ImageJ. Average intensities of regions of interest were quantified for each frame, and dynamic fluorescence signals were normalized to baseline values. The linear similarity (Pearson correlation) was calculated (Figure 4—figure supplement 1) between pairs of Ca2+ traces with a custom-made script in MATLAB (version 7.9.0.529 R2009b; MathWorks, Natick, MA, USA) and by using the mic2net toolbox (Smedler et al., 2014) (version 6.12; MathWorks). Calculating the pairwise correlation coefficients resulted in a correlation matrix that was converted to an adjacency matrix by applying a cut-off level. The cut-off level was selected by calculating the mean of the 99th percentile of correlation coefficients for a set of experiments with scrambled signals. Scrambling was performed by randomly translating all traces in the time-domain. The network structure was visualized by plotting a line between pairs of cells, where the color of the lines was proportionate to the correlation coefficient. This was plotted on top of an image of the standard deviations of the fluorescence over time per pixel. Connectivity was defined as the number of cell pairs with a correlation coefficient larger than the cut-off value divided by the total number of the pairs of cells. This provided a measure of the degree of connections within a network. Small-world parameter, mean shortest path length (λ) and mean clustering coefficient (σ) were calculated by using the MATLAB BGL library (http://www.mathworks.com/matlabcentral/fileexchange/10922) and compared to corresponding randomized networks. Many biological networks have a small-world structure, where the mean shortest path length is as short as in random networks and the mean clustering coefficient is higher. This signifies that the average number of nodes (for example, neural cells) that a signal has to pass is low, and that many of the nodes are connected in clusters (Watts and Strogatz, 1998). A small-world network structure creates the possibilities of regional specialization and efficient signal transfer, and is a common organization of networks within the brain (Telesford et al., 2011).

Data were further processed to produce graphs in OriginPro, version 9.1 (OriginLab Corporation, Northamptom, MA, USA). Time-lapse Ca2+ imaging time traces were normalized individually through ΔF/F0, where ΔF=F1−F0. F1 is the specific fluorescence intensity at a specific time point, and F0 is the average intensity of 30 s before and after F1.

A previously published toolbox was used for the frequency analysis of time traces (Uhlén, 2004).

Recordings of hypoglossal nerve activity were filtered (0.06-Hz low-pass), rectified and smoothed (1 s) (Talpalar et al., 2011) by using OriginPro (version 9.1, OriginLab Corporation, USA).

Statistics

Statistical analysis of paired comparisons was performed by Student’s t-test. Full factorial two-way ANOVA was performed when there was more than one independent variable or multiple observations. Both tests were two-sided. The compared data was of equal variance and normally distributed. All calculations for the statistical tests were conducted with JMP (v 11.1., SAS Institute Inc., Cary, NC, US). In all cases, p<0.05 was considered statistically significant. Data are presented as means ± SD. All data sets were compared less than 20 times, which is why no statistical corrections were made. As these experiments were expended to provide new descriptive data, no explicit power analysis was performed. Instead sample sizes similar to previous publications with similar methods were used. Details on the statistics are presented in Tables 9 and 10.

Table 9

Successful experiments behind representative images.

https://doi.org/10.7554/eLife.14170.067
FigurePanelNumber of experiments
1c25
2b
c
d
e
f
g
31
23
9
19
14
112
2 – S1112, 23, 12
33, 23, 10
27, 12, 11
22, 15, 8
15, 8, 8
2 – S25
2 – S3a
b
c
12
5
20
3a
b
5
5
3 – S1a
b
5
5
4b
c
d
g
h
12
13
8
9
5
4-S2840, 621, 456
5a16
5 – S1a6
5 – S2a
c
11
5
6a
b
c
d
e
f
g
h
i
j
9
8
8
3
8
6
7
8
6
7
7a
c
d
f
5, 4
9
12
5, 4
7 – S14, 4
8b
c
d
e
9
9
9
5
8 – S135, 22, 19
9, 9
5
9b
c
d
12
11
6
9 – S1315, 429, 192
9 – S218, 7
10a
b
7
7, 7 and 5
11a12, 4
11 – S1a
b
c
11
11
6
12a
c
d
e
f
5, 4
16
44
6
12
13a
c
f
6, 5
7, 5
41
13 – S1a
b
6
27
Table 10

Statistical details for presented figures.

https://doi.org/10.7554/eLife.14170.068
FigureTest usedExact p-valueDegrees of freedom&F/t/z/R value
1dFull factorial two-way ANOVA0.0084F=0.612, DFE=21
1eFull factorial two-way ANOVA0.0365F=0.284, DFE=21
1fFull factorial two-way ANOVA0.0157F=0.329, DFE=21
1gFull factorial two-way ANOVA0.0017
0.018
0.007
F=0.547
F=0.332
F=1.618
DFE=23
2 – S1dStudent’s t-test0.35
0.45
DF = 9, 19
4eStudent’s t-test0.63
0.76
0.91
DF=11, 12, 8
4fStudent’s t-test0.13, 0.34, 0.68
0.21, 0.28, 0.86
0.76, 0.76, 1.00
DF=11, 12, 8
4 – S2Student’s t-test0.43, 0.34, 0.12
0.11, 0.57, 0.19
DF=1916
4 – S2Paired t-test0.02, 0.02, 0.04
0.03, 0.02, 0.01
DF=1916
5bStudent’s t-test0.42, 0.51, 0.80DF=15
5cStudent’s t-test0.50, 0.62, 0.98DF=15
5 – S2aPaired t-test0.0029DF=6
5 – S2bPaired t-test0.041DF=10
5 – S2dPaired t-test0.03DF=4
6kPaired t-test0.03125
0.00391
0.3125
0.625
DF=23
6lPaired t-test0.01563
0.00781
0.28125
0.25
DF=23
6iPaired t-test0.492
0.331
0.390
0.390
DF=23
6mPaired t-test0.457
0.124
0.567
0.143
DF=23
6 – S1aPaired t-test0.15625
0.46094
0.0625
0.625
DF=23
6 – S1bPaired t-test0.15625
0.74219
0.15625
0.8125
DF=23
6 – S1cPaired t-test0.1246
0.07813
0.3125
0.8125
DF=23
7bPaired t-test0.018
0.034
0.047
0.079
0.132
0.084
0.028
0.063
0.067
0.012
0.077
0.90
DF= 4, 3, 7
7gPaired t-test0.95
0.51
DF=4, 3
7-S1Paired t-test0.51861
0.15558
0.69733
0.51508
0.36415
0.2433
DF=16
9eStudent’s t-test0.15, 0.51, 0.57
0.29, 0.061, 0.0081
DF=11, 10, 7
9fStudent’s t-test0.38, 0.51, 0.75
0.59, 0.66, 0.99
0.10, 0.10, 0.95
DF=11, 10, 7
9gPaired t-test0.43DF=3
9hPaired t-test0.0334DF=11
9iPaired t-test0.0141
0.0283
0.00475
DF=6, 5, 3
9 – S1Student’s t-test0.77, 0.51, 0.92
0.28, 0.07, 0.60
DF=935
9 – S1Paired t-test0.04, 0.04, 0.01
0.03, 0.02, 0.01
DF=935
9 – S2Paired t-test0.00507
0.5745
0.22731
0.68788
0.81018
0.66252
DF=16
10cPaired t-test0.53, 0.20, 0.61
0.009, 0.015, 0.041
0.023, 0.045, 0.035
0.01, 0.09, 0.14
DF=11, 7, 7, 9
10 – S1aPaired t-test0.153
0.0848
0.388
DF=6
10 – S1bPaired t-test0.59
0.43
DF=6
11bFull factorial two-way ANOVA0.418
0.0161
F=0.054
F=0.97
12bPaired t-test0.00038
0.28
DF= 6, 4
13bPaired t-test0.13, 0.56 0.16
0.24, 0.12, 0.012
DF=7
13dPaired t-test0.00046
0.87
DF=6, 4, 4, 3
13gPaired t-test0.4112
0.0001
DF=40
13 – S1cPaired t-test0.0125
0.00098
DF=5

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
    Prostaglandin E2 causes hypoventilation and apnea in newborn lambs
    1. FA Guerra
    2. RD Savich
    3. LD Wallen
    4. CH Lee
    5. RI Clyman
    6. FE Mauray
    7. JA Kitterman
    (1988)
    Journal of Applied Physiology 64:2160–2166.
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
    Current Protocols in Stem Cell Biology
    1. E Herlenius
    2. C Thonabulsombat
    3. D Forsberg
    4. J Jaderstad
    5. LM Jaderstad
    6. L Bjork
    7. P Olivius
    (2012)
    Unit 2D: Functional stem cell integration assessed by organotypic slice cultures, Current Protocols in Stem Cell Biology, Wiley, 10.1002/9780470151808.sc02d13s23.
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
    pH and Brain Function
    1. K Kaila
    2. BR Ransom
    (1998)
    New York: Wiley-Liss.
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
    Prostaglandins and control of breathing in newborn piglets
    1. WA Long
    (1988)
    Journal of Applied Physiology 64:409–418.
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
    Developmental changes in the hypoxic response of the hypoglossus respiratory motor output in vitro
    1. JM Ramirez
    2. UJ Quellmalz
    3. B Wilken
    (1997)
    Journal of Neurophysiology 78:383–392.
  66. 66
  67. 67
    Electrical coupling and excitatory synaptic transmission between rhythmogenic respiratory neurons in the preBötzinger complex
    1. JC Rekling
    2. XM Shao
    3. JL Feldman
    (2000)
    The Journal of Neuroscience 20:RC113.
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
  86. 86
  87. 87
  88. 88
  89. 89
  90. 90
  91. 91
  92. 92
  93. 93
  94. 94
  95. 95
    Coexpression of microsomal-type prostaglandin E synthase with cyclooxygenase-2 in brain endothelial cells of rats during endotoxin-induced fever
    1. K Yamagata
    2. K Matsumura
    3. W Inoue
    4. T Shiraki
    5. K Suzuki
    6. S Yasuda
    7. H Sugiura
    8. C Cao
    9. Y Watanabe
    10. S Kobayashi
    (2001)
    The Journal of Neuroscience 21:2669–2677.

Decision letter

  1. Jan-Marino Ramirez
    Reviewing Editor; Seattle Children's Research Institute and University of Washington, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your work entitled "CO2 induced release of PGE2 by glia increase inspiration and sighs through breathing brainstem neural networks" for consideration by eLife. Your article has been favorably evaluated by a Senior editor and three reviewers, one of whom, Jan-Marino Ramirez, is a member of our Board of Reviewing Editors.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

In the study by Forsberg et al., the authors combine in vivo, ex vivo and novel in vitro approaches to address the role of gap junctions and potential involvement of astrocytic release of prostaglandin E2 (PGE2) in respiratory control. PGE2 released in response to CO2 acts on prostaglandin EP3 receptors located in two main structures involved in respiratory rhythm generation, modulation and CO2 sensitivity: the pre-Bötzinger complex (preBötC) and the parafacial respiratory group/retrotrapezoid nucleus (pFRG/RTN). The authors propose that this PGE2-EP3R pathway is an essential component of the ventilatory response to CO2 and a vital modulator of respiratory activity, including sighing. The present results build upon previous results by Herlenius et al., showing that IL-1β impairs breathing during infection by inducing PGE2 release in the vicinity of brainstem respiratory networks (Hoffsteter et al. 2007), that inflammation via PGE2 and EP3R alters the ventilatory response to CO2 in brainstem spinal cord preparations (Siljehav et al., 2014), and other recent works on the role of PGE2 in the regulation of breathing pattern, in particular sigh frequency (Koch et al., 2015).

In order to characterize the pathway from brainstem astrocytes to the central pattern generator, the authors conducted plethysmographic studies in EP3R-knockout mice, which showed that PGE2 increases sighing activity via EP3R signaling and that the genetic ablation of EP3R reduces the hypercapnic response in vivo. In particular, they extended these experiments through electrophysiological studies in slice cultures of the brainstem with a functioning pFRG/RTN and preBötC. Furthermore, connectivity analysis of preBötC and pFRG/RTN respiratory networks (a very original aspect of the paper) showed that the hypercapnic response is dependent on gap junctions. The authors also used Ca2+ imaging and optogenetic silencing of EP3R-expressing cells to assess the role of this pathway in the hypercapnic response of the pFRG/RTN.

The reviewers were very impressed by the numerous techniques and novel approaches presented in this study. However, the manuscript is very complex and dense, and it tries to convey several different messages that in part become mixed up. For example, the thread of the Introduction oscillates between sighs, the role of astrocytes, inflammation, and inflammation-related respiratory disorders. A similar, general, concern also applies to the presentation of the results.

Another general concern is that the manuscript is lacking clear and concise specific hypotheses. The clarity of the manuscript would be greatly improved by clearly stating the hypothesis for a given set of experiments, and thus the conclusions of such experiments. At present, it is sometimes difficult to know how a given experiment relates to the overall conclusions. A similar concern relates to the figures. The reader must jump back and forth between different figures (and supplementary figures) to understand how the results relate to the specific and overall hypotheses of the manuscript.

It would vastly improve clarity to separate out analyses and hypotheses in the preBötC and those in pFRG/RTN. It would then be clearer why symmetrical analyses are not completed on each region.

Thus, in general, it will be critical to substantially re-organize the paper and figures, and make it more hypothesis-driven to bring it up to its full potential. Remember that the manuscript needs to be understood by a general audience.

Essential revisions:

1) The demonstration that it is actually the astrocytes that mediate the hypercapnic response is a minor and weaker part of the study. The study focuses more on the role of gap junctions than glia cells, and there is no experiment that actually tests the direct involvement of glia, or "astrocytic release" as stated in the Abstract. The role of glia is primarily inferred by prior work, e.g. by Gourine. The reviewers have no principle concern with the conclusion, and this possibility should be discussed, but the title, Introduction and Abstract should not imply that this study is about the role of glia. Along the same lines, the authors should avoid strong statements such as "PGE2 is released from astrocytes through CO2 sensitive connexin hemichannels". This is good for a discussion, but included into the Introduction (last paragraph) suggests that this will actually be demonstrated in the paper. Even with regards to the gap junctions. the authors did not really test gap junctions. The toxins used are not entirely specific. This should be discussed.

Along the same lines: A major conclusion of the manuscript is that PGE2 and ATP are released from astrocytes to alter pFRG/RTN responses to proton-independent increases in CO2 (Figure 12 and Discussion, sixth paragraph). However, this was not directly shown in the experiments conducted. The authors demonstrate changes after blocking gap junctions and blocking P2Rs alters the response, but this does not mean that astrocytes are releasing either PGE2 or ATP. The present results in context with previously published work does suggest this might be a logical conclusion, but alternative interpretations should be discussed.

2) The authors present a wealth of data using multiple different experimental methods, including a "breathing" brainstem slice culture. Several issues related to these slices were raised and are listed as separate points, below. Without addressing these concerns and being specific about the age of the slices in culture (i.e. 1, 2, or 3 weeks of culturing), it is difficult to critically evaluate the results. For clarity of the results, the reviewers suggest that it would be useful separate out data from the preBötC cultured slices and pFRG/RTN cultured slices, since the authors end up testing slightly different questions between the two preparations. This would help delineate which slice preparation was used for particular results. You will find related comments listed below.

3) Development of a "breathing" brainstem slice culture is very interesting and could be highly valuable to the field. However, the authors leave out a number of critical aspects with regards to the preparation that make the results difficult to evaluate. Firstly, it is unclear why such a preparation is necessary or what advantages it has compared to the classic rhythmic slice preparation. Acute measurements in the rhythmic slice preparation or a pFRG/RTN slice preparation would be just as advantageous for such experiments. Further, the authors are not clear about how long the slice culture is "breathing". No traces are shown to demonstrate baseline activity after 1, 2, or 3 weeks for the reader to see how activity changes with longer time in culture, and thus how the presented results might change with time. The authors do present data that the connectivity and number of active cells does not change with time (though the connectivity diagrams do appear to change over time especially when evaluating the raw numbers), but it is not clear how this would be observed in "breathing" or rhythmic/oscillatory behaviour. For example, is there any rhythmic activity after 3 weeks in culture in either the pFRG/RTN slices or the preBötC slices?

Additionally, no images are presented showing NK1 or other staining in the organotypic slices after three weeks, so it is difficult to decipher how neurons or astrocytes in culture have changed during this period (morphology, position, etc.). Since the conclusions in the study are based largely on data from such preparations, these are important methodological aspects to address. And since information is lacking in most figures and data presentation regarding the details of the slice preparation from which the data was obtained (i.e. 1, 2, or 3 weeks in culture), the reviewer assumes that the data from all weeks were collapsed, but without additional information, it is not clear whether such collapsing of data is appropriate. The authors also switch back and forth between saying "organotypic slice cultures", cultured cells, or slices. It would aid the reader if the authors used one terminology. Multiple types of preparations are used in these investigations and in the present format, it is difficult to follow which preparations are being used where and why such a preparation was necessary.

4) Related to the comment (3): There is a general issue with regards to the characterization of rhythmic activity in the two types of organotypic slices. Some of the specific concerns raised by the reviewers are as follows: The authors never spell out, whether they have population rhythms, and what kind of characteristics they have – even though they clearly have rhythms. The authors talk about "clustered activity", "clusters of high correlated activity". What do they mean with this? Only in the last few sentences of the subsection “Physiological measurements of brainstem respiratory activity demonstrate functional and responsive networks” they state that the preBötC remains rhythmic. The authors show something on Figure 3, but it doesn't look like rhythmicity. Yet, in the fourth paragraph of the aforementioned subsection, the authors imply that they have population rhythmicity that seems to even drive XII activity. Please try to be more concrete. Did these slices generate rhythmicity? What kind of properties do they find: frequency, amplitude, and also regularity? Do the calcium oscillations correspond to the "clustered" neuronal activities in frequency?

In the third paragraph of the subsection “Physiological measurements of brainstem respiratory activity demonstrate functional and responsive networks” – do the authors record from pacemaker neurons? This is potentially all very interesting, and if these slices show rhythmicity, the authors need to show traces of recordings, rather than activity from individual or pairs of neurons.

Figure 5 e.g. is beautiful, but do the pFRG/RTN slices show similar properties? Why is this not shown e.g. in Figure 7? Figure 4 is interesting, but why not show an expanded trace of rhythmicity? This is critical information in order to be able to assess what type of activity was characterized in this study. Did the pre-Bötzinger complex show activity that resembled sighs? Figure 11 shows some rhythmicity, but this should be expanded to see the characteristics. Given that this is the first report for these slices, a careful characterization is important. The authors might want to consider a reorganization of the manuscript: They should first characterize the population phenomenon, then report the clustered activity, correlations between neurons. Indeed, streamlining the messages, and not mixing up information would be critical to make this paper readable for a general readership.

5) Another general issue: the pFRG/RTN has been implicated (a) in CO2 sensing and (b) the neuronal control of active expiration. However, it is not entirely clear how these two functions work together, and whether the same neurons are responsible for active expiration and CO2 sensing. Phox2B is a good marker for the sensing aspect, but the rhythmicity observed in the pFRG/RTN slice may relate to "active expiration", and not necessarily to CO2 sensing. Clearly, the slice will also contain CO2 sensing neurons. Both possibilities are interesting, and both could play a role. The excitatory effect on the pFRG/RTN could imply that PGE2 induces active expiration, a possibility that has not been discussed in this paper. Discussing the role of active expiration should be considered, and the authors should also discuss the general issue of "rhythm generating function" versus "CO2 sensing", and the fact that this may not be related to the same neurons, but that PGE2 may affect both.

6) The reviewers have some reservations about the authors' claim that the PGE2-EP3R pathway is indispensable for the ventilatory response to hypercapnia. Specifically, the experimental support for this claim is not convincing (mainly in Figures 8 and 11). There are several problems in Figure 8 that may question whether the inhibition of gap junctions by 18-a-GA abolished the ventilatory response to CO2. On the one hand, Panel b (top trace) and c (left graph) seem to indicate that 18-α-GA abolishes the response to CO2. On the other hand, in Panel c, the CO2 responses are the same whether gap junctions are inhibited by 18-α-GA or not (GZA, i.e. the negative control for the gap junction inhibitors 18α-GA). In fact, control values (in air) in the middle and right graphs are very similar, and the values obtained for GZA+ CO2 and CO2+18-α-GA are also very similar. This apparent contradiction is reflected in the legend: it is first stated that "hypercarbic responses were abolished when gap junctions were inhibited by 18-α-GA", but later on, the legend states that "the hypercarbic response was attenuated but not completely reversed by 18-α-GA". Some details in this figure also need attention. DF/F0 in Panel b should not be expressed in "arbitrary units", as mentioned. It is in fact a dimensionless variable (there is nothing to add after "DF/F0" on the y-axis label). Also, the scales are different for the different traces in Panel b, which is misleading. The experimental basis of the abolition of the CO2 response in Figure 11 also presents some shortcomings. Panel a indicates a lack of response to CO2 after AH6809 treatment, but there is no experimental control showing the time course of frequency changes in the absence of AH6808. In Panel f, the top graph lacks a control (e. g. a return to normocapnia). As presented, it is unclear whether the response to CO2 was abolished, as implied by the authors, or whether the period of measurement was too short for the full CO2 response to appear. In the bottom graph, the light stimulus seems to stabilize the hypercapnic response, not to abolish it, which departs from the clear-cut return to baseline presented in Panel g, left graph. This raises a concern on the way data were averaged in Panel g.

Incidentally, the DF/F0 graph in Panel a contains a numerical scale on the y-axis, which should be the case for all similar graphs (with a larger font size).

Because of all these considerations it seems an exaggeration to conclude, that the PGE2-EP3R pathway is "a vital modulator of respiratory activity", if only because EP3R-/- mice (although obese) survive until adulthood.

7) The authors do not address why using a proton-independent CO2 stimulus was important to their studies, nor how they expect their studies to be altered if a proton-dependent stimulus was used. Since the majority of studies investigating hypercapnic responses has used acidification/pH changes as a stimulus (including the papers referenced in this manuscript), the authors should address this discrepancy. The authors need to address the rationale for choosing to investigate proton-independent changes in CO2 instead of changes in pH/H+, as is commonly done in the field. While CO2 may be a direct activator of cells, most literature to date has tested changes in pH. Additionally, many of the papers referenced in the manuscript in support of using proton-independent increased CO2, actually used pH to evoke changes. For example, Gourine et al. used changes in pH to examine involvement of ATP-P2Rs, yet the authors use these previous studies as rationale to look at the effects of proton-independent increased CO2 on the pFRG/RTN (Results, subsection “The pFRG/RTN respiratory region exhibited correlated network activity and retained CO2 sensitivity”). This discrepancy and rationale for using such a stimulus is not addressed in the present manuscript. Do the authors think that the same results would be obtained if experiments were conducted by altering pH? Why are the results proton-independent responses important? These questions are not addressed in the manuscript and would greatly improve the clarity and impact of the results.

8) The reviewers raised several questions regarding the experimental controls. Were the controls wild-type littermates of mutant mice (obtained by the crossing of heterozygous mice), and if so, were the heterozygotes also analyzed, as suggested by the mention that pups were analyzed after their genotype had been determined? If not, were the controls simply obtained from a C57Bl/6J strain? Control littermates are highly preferable, to control for date of birth, maternal care, litter size etc. These factors, also influence the maturation of CO2 sensitivity. It is indicated in the Methods that the EP3R gene was selectively deleted in knockout mice, as described in Fabre et al., 2001, but here, the authors used a 129 background.

9) The analysis of sighs is central to this study, and a very clear description of how sighs are defined and determined is thus essential. This point is ambiguous in the manuscript. In Methods, sighs are defined as "a breath with >50% increase in inspired volume compared to eupneic breaths", but there is no indication of how these reference eupneic breaths were determined. In Figure 1c and 1d, not all sighs (arrows) seem to fulfill these criteria. In the legend for Figure 1, sighs are defined differently, as "an inspiratory volume and respiratory cycle period increase followed by a 2-cycle pause". In fact, apnea does not occur systematically after a sigh and cannot be a defining characteristic. Besides, the trace in Figure 1b that illustrates the authors' definition does not display any post-sigh pause. In contrast, a defining characteristic of sighs is their biphasic pattern (see Toporikova et al., 2015 and the references quoted in her article). Without taking into account this biphasic pattern, above-average breaths may be confused with sighs, especially in a context of increased variability. There are also several minor concerns with Figure 1: sigh frequencies (Panel c) do not exemplify the mean values shown in Panel d; in Panels c, e and f, only relevant comparisons are necessary. In Figure 1—figure supplement 1, it is puzzling that there are 11 divisions between 100 and 200 on the y-axis. Also, with a 60% increase in Fr (compared to WT) and a 30% increase in VT (compared to WT), the EP3R-/- mice only displayed a 20-30% increase in VE (compared to WT). Is this correct?

10) Figure 1 needs to be revised: Three comments:

A) It is important to include "Figure 1—figure supplement 1" into the "final Figure 1". The supplementary figure contains an important message that is not contained in the Figure 1.

B) The traces in Figure 1f are not very helpful, as they contain lots of "artifacts" or deflections that are not respiratory, and could be mistaken by a non-specialist with sighs. I am very familiar with such traces and can detect sighs – but the general reader isn't an expert. Either, find other examples, or just show the graphs, which would also make it easier to add the supplementary figure.

C) Indeed, given that the sigh plays such an important part of the paper, there should be a figure of a sigh, and how the authors characterize sighs. Again, for the general reader this will not be clear from the figures presented. This comment relates also to the general comment (9) on the sigh.

11) Figure 12, has a very concrete model. It looks nice and is making their point. But, I am not convinced that they actually showed the differential glia/neuron modulation in preBötC and RTN/pFRG.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "CO2-evoked release of PGE2 modulates sighs and inspiration as demonstrated in brainstem organotypic culture" for further consideration at eLife. Your revised article has been favorably evaluated by a Senior editor, a Reviewing editor, and two reviewers.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

I would like to encourage you to cite and discuss the following study:

Organotypic slice cultures containing the preBötzinger complex generate respiratory-like rhythms.

Phillips WS, Herly M, Del Negro CA, Rekling JC.

J Neurophysiol. 2016 Feb 1;115(2):1063-70. doi: 10.1152/jn.00904.2015. Epub 2015 Dec 9.

PMID: 26655824

The publication of the organotypic slice does not diminish the impact of your study, but it would be important to acknowledge their study and also relate the activity that they report to your findings.

https://doi.org/10.7554/eLife.14170.070

Author response

The reviewers were very impressed by the numerous techniques and novel approaches presented in this study. However, the manuscript is very complex and dense, and it tries to convey several different messages that in part become mixed up. For example, the thread of the Introduction oscillates between sighs, the role of astrocytes, inflammation, and inflammation-related respiratory disorders. A similar, general, concern also applies to the presentation of the results.

[…]

It would vastly improve clarity to separate out analyses and hypotheses in the preBötC and those in pFRG/RTN. It would then be clearer why symmetrical analyses are not completed on each region.

Thus, in general, it will be critical to substantially re-organize the paper and figures, and make it more hypothesis-driven to bring it up to its full potential. Remember that the manuscript needs to be understood by a general audience.

We agree that the manuscript and the data are complex and the thread was difficult to follow in the original version. In accordance with your advice, we have now reorganized the manuscript and figures. The hypothesis for each set of experiments is now more clearly stated. We have also reorganized the results so that they are now divided into three themes. After the introduction to the main issues from the in vivo data, we first characterize the new organotypic brainstem slice culture system we have developed and use this to investigate the mechanisms behind the results seen in vivo. Since those experiments did not answer our main questions fully, we have now created a novel brainstem slice culture for the pFRG/RTN and characterized it as well. Finally, in the third part, we used the pFRG/RTN culture to investigate the role of PGE2 in chemosensitivity. Consequently, we have split some figures and rearranged them to follow the new structure of the results.

Essential revisions:

1) The demonstration that it is actually the astrocytes that mediate the hypercapnic response is a minor and weaker part of the study. The study focuses more on the role of gap junctions than glia cells, and there is no experiment that actually tests the direct involvement of glia, or "astrocytic release" as stated in the Abstract. The role of glia is primarily inferred by prior work, e.g. by Gourine. The reviewers have no principle concern with the conclusion, and this possibility should be discussed, but the title, Introduction and Abstract should not imply that this study is about the role of glia. Along the same lines, the authors should avoid strong statements such as "PGE2 is released from astrocytes through CO2 sensitive connexin hemichannels". This is good for a discussion, but included into the Introduction (last paragraph) suggests that this will actually be demonstrated in the paper. Even with regards to the gap junctions. the authors did not really test gap junctions. The toxins used are not entirely specific. This should be discussed.

Along the same lines: A major conclusion of the manuscript is that PGE2 and ATP are released from astrocytes to alter pFRG/RTN responses to proton-independent increases in CO2 (Figure 12 and Discussion, sixth paragraph). However, this was not directly shown in the experiments conducted. The authors demonstrate changes after blocking gap junctions and blocking P2Rs alters the response, but this does not mean that astrocytes are releasing either PGE2 or ATP. The present results in context with previously published work does suggest this might be a logical conclusion, but alternative interpretations should be discussed.

We agree with your statement that our data do not show the direct release of PGE2 (or ATP) from astrocytes. We have rephrased these conclusions throughout the manuscript and removed the statement that PGE2 is released by glia from the title, Abstract, and Introduction and have clarified that this is still a working hypothesis. In the fourteenth paragraph of the Discussion, we discuss the issue further and describe how astrocytes are known to release ATP, for example, and that it would be possible for them to release the similar small molecule PGE2 as well. Notably, we show that astrocytes lining the ventral border express mPGEs-1 (which produces PGE2). However, as you correctly point out, we have not shown directly that the PGE2 released during exposure to elevated CO2 originates from astrocytes. This is a remaining issue that needs to be resolved. Sources like endothelial cells and microglia should be investigated in future studies.

Further, we are aware of the problem with the gap junction inhibitors. However, they are commonly used for this purpose and we use both vehicle alone and structurally similar substances as controls. In the fourth paragraph of the Discussion, we discuss the limitations of gap junction pharmacological tools and that the results concerning gap junction function should be interpreted carefully.

We have also removed statements that would draw a line between proton-independent and -dependent mechanisms. In our experimental setup, extracellular pH remained stable during hypercapnic exposure but we have not controlled or monitored intracellular pH. Thus, our present experiments and data do not distinguish between the two alternatives.

2) The authors present a wealth of data using multiple different experimental methods, including a "breathing" brainstem slice culture. Several issues related to these slices were raised and are listed as separate points, below. Without addressing these concerns and being specific about the age of the slices in culture (i.e. 1, 2, or 3 weeks of culturing), it is difficult to critically evaluate the results. For clarity of the results, the reviewers suggest that it would be useful separate out data from the preBötC cultured slices and pFRG/RTN cultured slices, since the authors end up testing slightly different questions between the two preparations. This would help delineate which slice preparation was used for particular results. You will find related comments listed below.

Information about the age of the cultures during the experiment is very important, and we understand that this has raised concerns. Throughout the revised manuscript, we have specified how long the slices were cultivated before the experiment was performed, and we have also provided more detailed data on the behavior of the brainstem slice culture at different ages. This is presented in Figure 2—figure supplement 1, Figure 4—figure supplement 2, Figure 5—figure supplement 1, and Figure 9—figure supplement 1. In Figure 4—figure supplement 2 and Figure 9—figure supplement 1 we also display the calcium signaling frequency and variability at the different culture ages. In general, all the experiments in which we characterize the model, i.e., survival, immunohistochemical analysis of protein expression, calcium activity, network structure, and testing for the preservation of hypercarbic responses, used slices 1, 2, or 3 weeks of age. The more detailed experiments concerning inflammation and hypercarbic responses were done using 1-week-old cultures.

As we describe above, we have separated the preBötC from the pFRG/RTN data.

3) Development of a "breathing" brainstem slice culture is very interesting and could be highly valuable to the field. However, the authors leave out a number of critical aspects with regards to the preparation that make the results difficult to evaluate. Firstly, it is unclear why such a preparation is necessary or what advantages it has compared to the classic rhythmic slice preparation. Acute measurements in the rhythmic slice preparation or a pFRG/RTN slice preparation would be just as advantageous for such experiments. Further, the authors are not clear about how long the slice culture is "breathing". No traces are shown to demonstrate baseline activity after 1, 2, or 3 weeks for the reader to see how activity changes with longer time in culture, and thus how the presented results might change with time. The authors do present data that the connectivity and number of active cells does not change with time (though the connectivity diagrams do appear to change over time especially when evaluating the raw numbers), but it is not clear how this would be observed in "breathing" or rhythmic/oscillatory behaviour. For example, is there any rhythmic activity after 3 weeks in culture in either the pFRG/RTN slices or the preBötC slices?

Additionally, no images are presented showing NK1 or other staining in the organotypic slices after three weeks, so it is difficult to decipher how neurons or astrocytes in culture have changed during this period (morphology, position, etc.). Since the conclusions in the study are based largely on data from such preparations, these are important methodological aspects to address. And since information is lacking in most figures and data presentation regarding the details of the slice preparation from which the data was obtained (i.e. 1, 2, or 3 weeks in culture), the reviewer assumes that the data from all weeks were collapsed, but without additional information, it is not clear whether such collapsing of data is appropriate. The authors also switch back and forth between saying "organotypic slice cultures", cultured cells, or slices. It would aid the reader if the authors used one terminology. Multiple types of preparations are used in these investigations and in the present format, it is difficult to follow which preparations are being used where and why such a preparation was necessary.

We are happy to hear that the development of a breathing brainstem organotypic culture might be highly valuable to the present field of research. In the short introduction to the first theme (Results), we point out that there is a need for a model system that provides the ability to perform long-term studies on respiratory centers. In the second paragraph of the Discussion, we now provide a more in-depth discussion on the pros and cons of our brainstem slice culture. In short, the culture provides the possibility of studying long-term drug exposure (hours to days), which we are currently conducting in parallel in studies that have not yet been finalized, and viral transduction in vitro (not possible in “acute slices”), which we have assessed for the optogenetic experiments in this study. Further, the slice cultures give the user somewhat more flexibility and significantly reduces the number of procedures otherwise needed to be performed on live animals.

To better characterize and clarify how the baseline activity of the cultures changes over time, we have added several figures and data. In Figure 4—figure supplement 2 and Figure 9—figure supplement 1 we have added traces of NK1R-expressing neurons’ activity during TTX inhibition at 7, 14, and 21 days in vitro. We have not detected differences among slices of different ages with any of the parameters we have analyzed. In Figure 5—figure supplement 1, we show rhythmic hypoglossal nerve output that is stable for over 2 hours of recording after 3 weeks of cultivation.

We have now shown the expression of the different neural markers after 1, 2, and 3 weeks of slice cultivation in Figure 2—figure supplement 2. Throughout the whole manuscript, we have better described the settings for each experiment, including cultivation stage. For consistency, we have used “brainstem slice culture” for general terminology.

The current characterization better presents the novel model system’s behavior over time. The present manuscript already contains a wealth of novel data, techniques, and approaches, as mentioned by the reviewers. Nonetheless, an even more detailed characterization will surely emerge in future studies. We hope that the novel approaches presented in the current manuscript will help to shed light on several of the remaining black boxes of how inspiration is generated and controlled.

4) Related to the comment (3): There is a general issue with regards to the characterization of rhythmic activity in the two types of organotypic slices. Some of the specific concerns raised by the reviewers are as follows: The authors never spell out, whether they have population rhythms, and what kind of characteristics they have – even though they clearly have rhythms. The authors talk about "clustered activity", "clusters of high correlated activity". What do they mean with this? Only in the last few sentences of the subsection “Physiological measurements of brainstem respiratory activity demonstrate functional and responsive networks” they state that the preBötC remains rhythmic. The authors show something on Figure 3, but it doesn't look like rhythmicity. Yet, in the fourth paragraph of the aforementioned subsection, the authors imply that they have population rhythmicity that seems to even drive XII activity. Please try to be more concrete. Did these slices generate rhythmicity? What kind of properties do they find: frequency, amplitude, and also regularity? Do the calcium oscillations correspond to the "clustered" neuronal activities in frequency?

In the third paragraph of the subsection “Physiological measurements of brainstem respiratory activity demonstrate functional and responsive networks” – do the authors record from pacemaker neurons? This is potentially all very interesting, and if these slices show rhythmicity, the authors need to show traces of recordings, rather than activity from individual or pairs of neurons.

Figure 5 e.g. is beautiful, but do the pFRG/RTN slices show similar properties? Why is this not shown e.g. in Figure 7? Figure 4 is interesting, but why not show an expanded trace of rhythmicity? This is critical information in order to be able to assess what type of activity was characterized in this study. Did the pre-Bötzinger complex show activity that resembled sighs? Figure 11 shows some rhythmicity, but this should be expanded to see the characteristics. Given that this is the first report for these slices, a careful characterization is important. The authors might want to consider a reorganization of the manuscript: They should first characterize the population phenomenon, then report the clustered activity, correlations between neurons. Indeed, streamlining the messages, and not mixing up information would be critical to make this paper readable for a general readership.

We hope our revised version better addresses your concerns on rhythmicity. We have also added more data to provide a more detailed description of the rhythmic activity (see above). The slices do indeed generate rhythmicity. Our results are based on calcium oscillations within the preBötC and extracellular measurements of the electrical hypoglossal nerve and nucleus activity. Results indicate the presence of NK1R+ neuronal pacemaker activity. Their rhythmic Ca2+ signals during inhibition of synaptic transmission remained, however with a lower frequency and higher coefficient of variation (subsection “Physiological measurements of brainstem respiratory activity demonstrate functional and responsive networks”, fifth paragraph). We have also added a display of a longer extracellular recording from the hypoglossal nucleus, showing activity after 3 weeks, as well as graphs that show frequency and regularity at 1, 2, and 3 weeks (Figure 5 andFigure 5—figure supplement 1).

Considering the clustered activity, we refer to cells that were shown to be connected in the correlation analysis. We have added Figure 4—figure supplement 1, for clarification and to make this concept easier to grasp. Also, calcium oscillation frequency does indeed correspond to motor output frequency.

Regarding sigh activity, we have noticed increases in the calcium signal amplitude and length after application of PGE2. This is discussed in the second paragraph of the subsection “PGE2 modulates preBötC activity”. We have not, however, identified sigh-related calcium activity in preBötC control recordings.

Concerning the pFRG/RTN output, to our knowledge, there is no established method to measure such electrical nerve output from slices, and hence such a technique is not established in our lab. Respiration-related output has been reported to emerge through the facial nerve. However, in those studies, preBötC and other respiratory centers were preserved in the specimen examined.

5) Another general issue: the pFRG/RTN has been implicated (a) in CO2 sensing and (b) the neuronal control of active expiration. However, it is not entirely clear how these two functions work together, and whether the same neurons are responsible for active expiration and CO2 sensing. Phox2B is a good marker for the sensing aspect, but the rhythmicity observed in the pFRG/RTN slice may relate to "active expiration", and not necessarily to CO2 sensing. Clearly, the slice will also contain CO2 sensing neurons. Both possibilities are interesting, and both could play a role. The excitatory effect on the pFRG/RTN could imply that PGE2 induces active expiration, a possibility that has not been discussed in this paper. Discussing the role of active expiration should be considered, and the authors should also discuss the general issue of "rhythm generating function" versus "CO2 sensing", and the fact that this may not be related to the same neurons, but that PGE2 may affect both.

Active expiration has not been the primary focus of this manuscript, but we have added the concept in the eleventh paragraph of the Discussion. It is possible that PGE2 stimulates both chemosensitive neurons and neurons important for active expiration. Such neurons could also overlap. Future studies should aim to investigate whether PGE2 also affects active expiration.

6) The reviewers have some reservations about the authors' claim that the PGE2-EP3R pathway is indispensable for the ventilatory response to hypercapnia. Specifically, the experimental support for this claim is not convincing (mainly in Figures 8 and 11). There are several problems in Figure 8 that may question whether the inhibition of gap junctions by 18-a-GA abolished the ventilatory response to CO2. On the one hand, Panel b (top trace) and c (left graph) seem to indicate that 18-α-GA abolishes the response to CO2. On the other hand, in Panel c, the CO2 responses are the same whether gap junctions are inhibited by 18-α-GA or not (GZA, i.e. the negative control for the gap junction inhibitors 18α-GA). In fact, control values (in air) in the middle and right graphs are very similar, and the values obtained for GZA+ CO2 and CO2+18-α-GA are also very similar. This apparent contradiction is reflected in the legend: it is first stated that "hypercarbic responses were abolished when gap junctions were inhibited by 18-α-GA", but later on, the legend states that "the hypercarbic response was attenuated but not completely reversed by 18-α-GA". Some details in this figure also need attention. DF/F0 in Panel b should not be expressed in "arbitrary units", as mentioned. It is in fact a dimensionless variable (there is nothing to add after "DF/F0" on the y-axis label). Also, the scales are different for the different traces in Panel b, which is misleading. The experimental basis of the abolition of the CO2 response in Figure 11 also presents some shortcomings. Panel a indicates a lack of response to CO2 after AH6809 treatment, but there is no experimental control showing the time course of frequency changes in the absence of AH6808. In Panel f, the top graph lacks a control (e. g. a return to normocapnia). As presented, it is unclear whether the response to CO2 was abolished, as implied by the authors, or whether the period of measurement was too short for the full CO2 response to appear. In the bottom graph, the light stimulus seems to stabilize the hypercapnic response, not to abolish it, which departs from the clear-cut return to baseline presented in Panel g, left graph. This raises a concern on the way data were averaged in Panel g.

Incidentally, the DF/F0 graph in Panel a contains a numerical scale on the y-axis, which should be the case for all similar graphs (with a larger font size). Because of all these considerations it seems an exaggeration to conclude, that the PGE2-EP3R pathway is "a vital modulator of respiratory activity", if only because EP3R-/- mice (although obese) survive until adulthood.

We agree that the PGE2-EP3R pathway is not indispensable, but our results show that it is important, given the attenuated response in EP3R knockout mice. Therefore, we have replaced “indispensable” and “vital” with “important” in the text.

The comment on Figure 8 is an interesting point. Indeed, there is no significant difference between frequencies during treatment with CO2+18-α-GA and CO2+GZA. However, the dynamics are different. When gap junctions are inhibited by 18-α-GA, no CO2 response is seen. This is not the case with GZA, as CO2 increases signaling frequency during GZA treatment. On the other hand, 18-α-GA is not able to completely reverse the CO2 response, supporting our theory that it is mainly pre-made PGE2 that is released. However, the reduction that 18-α-GA causes, indicates that gap junctions are still important to maintain the CO2 response, even if they are not exclusively responsible for this function. GZA is not able to reduce an already-initiated response. We have adjusted this in the figure (which is now Figure 10) to better illustrate this.

The DF/F0 panels have been adjusted throughout the manuscript according to your suggestion. We have also added the numerical scale to all relevant figures. We cannot, however, see how the scales are different in Figure 8b (now Figure 10b). The middle (GZA) trace is somewhat noisier compared to the other two. Is this what you are referring to?

A control recording has been added in Figure 13a for comparison, and a graph showing average frequencies in these experiments has been added to Figure 13b. A trace showing the return to normocapnia has been added to Figure 13f, as you have suggested. Also, a cell trace (from the same experimental setting) that better represents the average data has been used for the bottom graph in Figure 13f.

7) The authors do not address why using a proton-independent CO2 stimulus was important to their studies, nor how they expect their studies to be altered if a proton-dependent stimulus was used. Since the majority of studies investigating hypercapnic responses has used acidification/pH changes as a stimulus (including the papers referenced in this manuscript), the authors should address this discrepancy. The authors need to address the rationale for choosing to investigate proton-independent changes in CO2 instead of changes in pH/H+, as is commonly done in the field. While CO2 may be a direct activator of cells, most literature to date has tested changes in pH. Additionally, many of the papers referenced in the manuscript in support of using proton-independent increased CO2, actually used pH to evoke changes. For example, Gourine et al. used changes in pH to examine involvement of ATP-P2Rs, yet the authors use these previous studies as rationale to look at the effects of proton-independent increased CO2 on the pFRG/RTN (Results, subsection “The pFRG/RTN respiratory region exhibited correlated network activity and retained CO2 sensitivity”). This discrepancy and rationale for using such a stimulus is not addressed in the present manuscript. Do the authors think that the same results would be obtained if experiments were conducted by altering pH? Why are the results proton-independent responses important? These questions are not addressed in the manuscript and would greatly improve the clarity and impact of the results.

As stated above, our experiments do not distinguish between proton-dependent and -independent CO2 responses. With regard to intracellular pH, our experimental designs will have an effect e.g. (Ruusuvuori, Kirilkin et al. 2010).

8) The reviewers raised several questions regarding the experimental controls. Were the controls wild-type littermates of mutant mice (obtained by the crossing of heterozygous mice), and if so, were the heterozygotes also analyzed, as suggested by the mention that pups were analyzed after their genotype had been determined? If not, were the controls simply obtained from a C57Bl/6J strain? Control littermates are highly preferable, to control for date of birth, maternal care, litter size etc. These factors, also influence the maturation of CO2 sensitivity. It is indicated in the Methods that the EP3R gene was selectively deleted in knockout mice, as described in Fabre et al., 2001, but here, the authors used a 129 background.

Unfortunately, we do not have the ability to use littermate controls, as the strain is the homozygotic EP3R knockout obtained from the Anders Blomqvist research group at Linköping University, Sweden. Genotyping was performed to confirm the genetic homogeneity of the knockout line. These mice do indeed have a C57Bl/6J background and are regularly backcrossed with C57Bl/6J, which is why such animals, bred in the same facility, were used as wild-type controls. The reference by Fabre et al. simply describes the technique used, though in a different strain from the one used here. A clarification has been added to Methods, subsection “Subjects”, first paragraph.

9) The analysis of sighs is central to this study, and a very clear description of how sighs are defined and determined is thus essential. This point is ambiguous in the manuscript. In Methods, sighs are defined as "a breath with >50% increase in inspired volume compared to eupneic breaths", but there is no indication of how these reference eupneic breaths were determined. In Figure 1c and 1d, not all sighs (arrows) seem to fulfill these criteria. In the legend for Figure 1, sighs are defined differently, as "an inspiratory volume and respiratory cycle period increase followed by a 2-cycle pause". In fact, apnea does not occur systematically after a sigh and cannot be a defining characteristic. Besides, the trace in Figure 1b that illustrates the authors' definition does not display any post-sigh pause. In contrast, a defining characteristic of sighs is their biphasic pattern (see Toporikova et al., 2015 and the references quoted in her article). Without taking into account this biphasic pattern, above-average breaths may be confused with sighs, especially in a context of increased variability. There are also several minor concerns with Figure 1: sigh frequencies (Panel c) do not exemplify the mean values shown in Panel d; in Panels c, e and f, only relevant comparisons are necessary. In Figure 1—figure supplement 1, it is puzzling that there are 11 divisions between 100 and 200 on the y-axis. Also, with a 60% increase in Fr (compared to WT) and a 30% increase in VT (compared to WT), the EP3R-/- mice only displayed a 20-30% increase in VE (compared to WT). Is this correct?

We have better defined the sigh (as suggested in Toporikova 2015) on page 4 and in the Materials and methods section. In Figure 1b, we now more illustratively show its characteristics. In Figure 1c, we selected a different period from the same experimental setting that better represents the mean values presented in Figure 1d. In Figure 1e, there was a shift in the graph, but, after rendering it again, this has been solved. The differences in the figure supplement (which is now incorporated into the main Figure 1) as you observantly noted was due to a mix-up in the graphs. The Fq graph was the VE graph and vice versa. This embarrassing mistake has been corrected, and we sincerely apologize for this. The statements in the Results text, Table 1, and the legend were correct. The traces in the previous Figure 1F have been removed in accordance with your suggestion.

10) Figure 1 needs to be revised: Three comments:

A) It is important to include "Figure 1—figure supplement 1" into the "final Figure 1". The supplementary figure contains an important message that is not contained in the Figure 1.

B) The traces in Figure 1F are not very helpful, as they contain lots of "artifacts" or deflections that are not respiratory, and could be mistaken by a non-specialist with sighs. I am very familiar with such traces and can detect sighs – but the general reader isn't an expert. Either, find other examples, or just show the graphs, which would also make it easier to add the supplementary figure.

C) Indeed, given that the sigh plays such an important part of the paper, there should be a figure of a sigh, and how the authors characterize sighs. Again, for the general reader this will not be clear from the figures presented. This comment relates also to the general comment (9) on the sigh.

Please see the answer to comment (9).

11) Figure 12, has a very concrete model. It looks nice and is making their point. But, I am not convinced that they actually showed the differential glia/neuron modulation in preBötC and RTN/pFRG.

Thank you. The current Figure is now 14. We agree that we have not yet shown directly the differential glia/neuron modulation and have therefore added a question mark at the astrocyte to better illustrate that this is a possible mechanism that remains to be addressed in future work.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

I would like to encourage you to cite and discuss the following study:

Organotypic slice cultures containing the preBötzinger complex generate respiratory-like rhythms.

Phillips WS, Herly M, Del Negro CA, Rekling JC.

J Neurophysiol. 2016 Feb 1;115(2):1063-70. doi: 10.1152/jn.00904.2015. Epub 2015 Dec 9.

PMID: 26655824

The publication of the organotypic slice does not diminish the impact of your study, but it would be important to acknowledge their study and also relate the activity that they report to your findings.

We have thoroughly read this study and find it interesting and intriguing as it both confirms and complements our data. We refer to it in the second paragraph of the Discussion.

https://doi.org/10.7554/eLife.14170.071

Article and author information

Author details

  1. David Forsberg

    1. Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden
    2. Karolinska University Hospital, Stockholm, Sweden
    Contribution
    DF, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4719-2201
  2. Zachi Horn

    1. Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden
    2. Karolinska University Hospital, Stockholm, Sweden
    Contribution
    ZH, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Contributed equally with
    Evangelia Tserga
    Competing interests
    No competing interests declared.
  3. Evangelia Tserga

    1. Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden
    2. Karolinska University Hospital, Stockholm, Sweden
    Contribution
    ET, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Contributed equally with
    Zachi Horn
    Competing interests
    No competing interests declared.
  4. Erik Smedler

    Department of Medical Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden
    Contribution
    ES, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4609-3620
  5. Gilad Silberberg

    Department of Neuroscience, Karolinska Institutet, Stockholm, Sweden
    Contribution
    GS, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9964-505X
  6. Yuri Shvarev

    1. Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden
    2. Karolinska University Hospital, Stockholm, Sweden
    Contribution
    YS, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6622-1453
  7. Kai Kaila

    Department of Biosciences and Neuroscience Center, University of Helsinki, Helsinki, Finland
    Contribution
    KK, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0668-5955
  8. Per Uhlén

    Department of Medical Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden
    Contribution
    PU, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1446-1062
  9. Eric Herlenius

    1. Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden
    2. Karolinska University Hospital, Stockholm, Sweden
    Contribution
    EH, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    eric.herlenius@ki.se
    Competing interests
    EH: employed at the Karolinska University Hospital and the Karolinska Institutet and is a coinventor of a patent application regarding biomarkers and their relation to breathing disorders, WO2009063226.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6859-0620

Funding

Karolinska Institutet (Graduate MD PhD student fellowship)

  • David Forsberg

Vetenskapsrådet (2010-4392)

  • Per Uhlén

Hjärnfonden (FO2014+0220)

  • Per Uhlén

Vetenskapsrådet (2013-3189)

  • Per Uhlén

Vetenskapsrådet (Senior Clinical researcher 6-year position, 2008-5829)

  • Eric Herlenius

VINNOVA (Future Health Innovation grant, 2010-00534)

  • Eric Herlenius

Stockholms Läns Landsting (2011-0095)

  • Eric Herlenius

Hjärnfonden (FO2011-008)

  • Eric Herlenius

The Swedish National Heart and Lung foundation (20120373)

  • Eric Herlenius

Vetenskapsrådet (2009-3724)

  • Eric Herlenius

Knut och Alice Wallenbergs Stiftelse (Senior clinical researcher award and research grant, 102179)

  • Eric Herlenius

Stockholms Läns Landsting (2012-0465)

  • Eric Herlenius

Stockholms Läns Landsting (2014-0011)

  • Eric Herlenius

Hjärnfonden (FO2012-0036)

  • Eric Herlenius

Hjärnfonden (FO2015-0020)

  • Eric Herlenius

The Swedish National Heart and Lung foundation (20150558)

  • Eric Herlenius

the Freemasons’ Children’s House (2015)

  • Eric Herlenius

the Axel Tielman Foundation (2015-00220)

  • Eric Herlenius

The Fraenckel Foundation (FRAE0018)

  • Eric Herlenius

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Markus Kruusmägi, Josephine Forsberg, Ruth Detlofsson, Dorina Ujvari, David Lagman, Torkel Mattesson and Lars Björk for their technical assistance; John Hayes for ImageJ scripts and advice; Marie Carlén and Thomas Ringstedt for their advice and discussion concerning optogenetics; Anders Blomqvist and Unn Örtegren Kugelberg (Linköping University, Sweden) for providing Ptger3-/- mice; and Ed Boyden for providing halorhodopsin-57. This study was supported by the Swedish Research Council, the Stockholm County Council, the Karolinska Institutet, and grants from the VINNOVA, M & M Wallenberg, Fraenkel, Axel Tielman, Freemasons Children’s House and Swedish National Heart and Lung Foundations.

Ethics

Animal experimentation: The studies were performed in strict accordance with European Community Guidelines and protocols approved by the regional ethic committee (Permit numbers: N247/13, N265/14b & N185/15).

Reviewing Editor

  1. Jan-Marino Ramirez, Seattle Children's Research Institute and University of Washington, United States

Publication history

  1. Received: January 3, 2016
  2. Accepted: June 21, 2016
  3. Accepted Manuscript published: July 5, 2016 (version 1)
  4. Version of Record published: August 4, 2016 (version 2)

Copyright

© 2016, Forsberg et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,964
    Page views
  • 446
    Downloads
  • 8
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Computational and Systems Biology
    2. Neuroscience
    Daniel B Dorman et al.
    Research Article Updated
    1. Computational and Systems Biology
    2. Neuroscience
    Dilawar Singh, Upinder Singh Bhalla
    Research Article