Time and dose-dependent impairment of neonatal respiratory motor activity after systemic inflammation

Abstract

Neonatal respiratory impairment during infection is common, yet its effects on respiratory neural circuitry are not fully understood. We hypothesized that the timing and severity of systemic inflammation is positively correlated with impairment in neonatal respiratory activity. To test this, we evaluated time- and dose-dependent impairment of in vitro fictive respiratory activity. Systemic inflammation (induced by lipopolysaccharide, LPS, 5 mg/kg, i.p.) impaired burst amplitude during the early (1h) inflammatory response. The greatest impairment in respiratory activity (decreased amplitude, frequency, and increased rhythm disturbances) occurred during the peak (3h) inflammatory response in brainstem-spinal cord preparations. Surprisingly, isolated medullary respiratory circuitry within rhythmic slices showed decreased baseline frequency and delayed onset of rhythm only after higher systemic inflammation (LPS 10 mg/kg) early in the inflammatory response (1h), with no impairments at the peak inflammatory response (3h). Thus, different components of neonatal respiratory circuitry have differential temporal and dose sensitivities to systemic inflammation, creating multiple windows of vulnerability for neonates after systemic inflammation.

1. Introduction

The respiratory control system is immature at birth (Allan & Greer, 1997; Greer & Funk, 2005; Ramirez et al., 1996) and neonates are fraught with inflammatory stimuli (Osrin, et al., 2004; Skogstrand et al., 2008; Stoll et al., 2002, 2004); thus, there is considerable potential for systemic and (indirectly) CNS inflammation to undermine proper development of vital respiratory circuitry controlling breathing. Low birth weight, associated with prematurity, and neonatal infection are among the top ten leading causes of infant death. Premature infants are also more vulnerable to acquiring an infection (Aly et al., 2005; Burgner et al., 1996; Seo et al., 1992) due to increased duration of hospital stays (Örtenstrand et al., 2010; Petrou et al., 2003) and an immature immune system (Ballow et al., 1986; Melville & Moss, 2013). Since neonatal infection is linked to respiratory dysfunction and impairment (Hofstetter et al., 2008; Palmer et al., 2002; Speer, 2006; Wu et al., 2008), we sought to better understand the interactions between early life inflammation and respiratory dysfunction.

While neonatal inflammation contributes to peripheral dysfunction early in life, such as bronchopulmonary dysplasia (Speer, 2006) and childhood asthma (Palmer et al., 2002; Wu et al., 2008), we are only beginning to understand the effects of neonatal inflammation on central networks controlling breathing. Premature infants have more apneas relative to term infants (Barrington & Finer, 1990; Daily et al., 1969; Miller et al., 1959; Schmidt et al., 2006), suggesting greater instability of central networks controlling breathing with immaturity. Further, neonatal inflammation augments apneas and hypopneas in extremely preterm infants compared with premature infants without infection (Hofstetter et al., 2008), suggesting a synergistic negative effect of neonatal inflammation on developing respiratory circuitry. We previously demonstrated gestational intermittent hypoxia induces neonatal inflammation and acutely impairs isolated central respiratory networks (Johnson et al., 2018). In contrast, intratracheal-evoked systemic inflammation using lipopolysaccharide (LPS, a component of the cell wall of gram-negative bacteria eliciting toll-like receptor 4-mediated systemic inflammation) increases respiratory frequency (Gresham et al., 2011), highlighting discrepancies in the effects of neonatal inflammation on control of breathing. Thus, the effects of neonatal inflammation on central respiratory circuitry are likely influenced by the timing of inflammation during development and the inflammation severity.

Clinically, neonatal infections range from mild (associated with apnea of prematurity, Bruhn et al., 1977) to severe (causing sepsis, Franciosi et al., 1973; Stoll et al., 2011), demonstrating the need to understand the effect of inflammation severity in neonates. In adults, severe systemic inflammation has dire consequences, where high doses (>20 mg/kg) (Ben-Shaul et al., 2001) and moderate-high doses (10 or 20 mg/kg) (Lew et al., 2013) of LPS increase mortality, highlighting severe consequences after adult inflammation. More moderate LPS doses (3 mg/kg, Vinit et al., 2011) and even low dose LPS (100 μg/kg) attenuates hypoxic ventilatory responses (Master et al., 2016) and impairs adult respiratory motor plasticity (Huxtable et al., 2013; Huxtable et al., 2015; Vinit et al., 2011). Thus, while the effects of inflammation in adults are dependent on the severity of inflammation, the effects in neonates are unclear. Here, we investigate the impact of LPS-induced (0.1–10 mg/kg) systemic inflammation on isolated, central respiratory control networks in the medulla and spinal cord.

The most commonly used model to evoke peripheral and indirectly central inflammation is LPS. LPS causes peripheral inflammation through activation of toll-like receptor 4 (TLR4) and recruitment of adaptor proteins (MyD88, TIRAP, TRIF, TRAM, and SARM) (Kawai et al., 2001; O’Neill & Bowie, 2007). This activation causes translocation of transcription factor NF-κB to the nucleus (Cordle et al., 1993; Müller et al., 1993), increasing cytokine gene and protein expression (Cho et al., 2003). Increased peripheral cytokines can directly cross the blood-brain barrier through active transport (Banks et al., 1989; Banks et al., 1991; Banks et al., 1994; Pan & Kastin, 2002) and activate vagal afferents (Balan et al., 2011; Buller, 2001) to induce neuroinflammation through a process involving activation of microglia (Henry et al., 2009; Sweitzer et al., 1999) and astrocytes (Sweitzer et al., 1999). Previous work in adults demonstrated the profile of inflammatory gene expression within the CNS is dependent both on the inflammatory stimulus type (LPS vs. nocturnal intermittent hypoxia) and timing (3 vs. 24h) (Huxtable et al., 2013; Huxtable et al., 2015). Further, CNS regional differences (medulla vs. spinal cord) in inflammatory gene expression after chronic intermittent hypoxia suggest distinct aspects of the respiratory circuitry respond differently to systemic inflammation (Smith et al., 2013). While we are just beginning to understand the effects of peripheral inflammation on adult respiratory-related regions, we know even less about the effects of peripheral inflammation on central respiratory regions in neonates. Low LPS doses (200 μg/kg) increase serum cytokine levels in neonates (McDonald et al., 2016); however, our understanding of the impact on central respiratory networks in response to peripheral inflammation remains incomplete.

Thus, we investigated the effects of neonatal inflammation during the inflammatory response on central respiratory control networks. We investigated the severity of inflammation induced by different LPS doses (0.1–10 mg/kg) on isolated central respiratory circuitry at 1h and 3h post-injection. We hypothesized respiratory-related activity would decrease with increasing inflammation severity, and would be most impaired during the peak inflammatory response. Understanding the influence of peripheral inflammation on the control of breathing will identify the role of inflammation in neonatal dysfunction or mortality, and lead to better therapeutic interventions to treat respiratory impairments associated with neonatal infections.

2. Materials and methods

All experimental procedures followed the National Institutes of Health Guide for the Care of Use of Laboratory Animals and were approved by the University of Oregon and University of Wisconsin-Madison Institutional Animal Care and Use Committees. Timed-pregnant Sprague-Dawley female rats were obtained from Charles River (colony R04, R08; Wilmington, MA, USA) or Envigo (colony 202A, 206; Haslett, MI, USA) between G15–19 and housed in standard rat cages with 12:12 hour light/dark cycle and ad libitum access to food and water. Neonatal rats (male and female) aged P0–P4 were used for brainstem-spinal cord (BSSC, also known as the en bloc in vitro preparation) experiments, and P0–P6 rats were used for rhythmic slice experiments. Since no obvious sex differences were found, the sexes were pooled.

2.1. Drugs and materials

A lipopolysaccharide (LPS, Sigma-Aldrich, Milwaukee, WI) stock solution (10 mg/ml in saline) was sonicated and frozen. Prior to use, LPS was sonicated and diluted in saline to final doses of 0.1, 1, 5, and 10 mg/kg. Since LPS administered systemically does not cross the blood-brain barrier (Banks & Robinson, 2010), the timing of LPS injections relative to commencement of dissections is critical as the peripheral inflammatory signal is communicated indirectly to the central nervous system. Common modes of transport of the peripheral inflammatory signal to the CNS include: activation of vagal afferents (Balan et al., 2011; Buller, 2001), active transport of cytokines across the blood-brain barrier (Banks et al., 1989; Banks et al., 1991; Banks et al., 1994; Pan & Kastin, 2002) and at circumventricular organs (Quan t al., 1999). To test the effects on the CNS after 1 or 3 hours of peripheral to central inflammatory signaling, dissections began 1 or 3 hours after LPS injection. Additionally, since tissue isolation and dissections were conducted at low temperatures (<17°C), enhancement of the inflammatory signals during dissections is expected to be minimal.

2.2. Experimental groups

To investigate the temporal effects of systemic inflammation on medullary and cervical spinal cord inflammatory gene expression, neonatal rats were assigned to six groups (n = 6 rats/group): 1) 1h saline, 2) 1h LPS (1 mg/kg), 3) 3h saline, 4) 3h LPS (1 mg/kg), 5) 24h saline, 6) 24h LPS (1 mg/kg).

To test the hypothesis that LPS-induced acute systemic inflammation decreases neonatal respiratory activity at 1h and 3h post-injection in BSSC preparations, neonatal rats were assigned to four groups at 1h: 1) saline (n = 14), 2) 1 mg/kg LPS (n = 11), 3) 5 mg/kg LPS (n = 5), 4) 10 mg/kg LPS (n = 6), and four groups at 3h: 1) saline (n = 14), 2) 0.1 mg/kg LPS (n = 21), 3) 1 mg/kg LPS (n = 17), 4) 10 mg/kg LPS (n = 10).

To test the hypothesis that LPS-induced acute systemic inflammation decreases respiratory activity in rhythmically active medullary slices at 1h and 3h post-injection, neonatal rat pups were assigned to three groups at 1h: 1) saline (n = 6), 2) 1 mg/kg LPS (n= 5), 3) 10 mg/kg LPS (n = 5), and three groups at 3h: 1) saline (n = 6), 2) 1 mg/kg LPS (n = 6), 3) 10 mg/kg LPS (n = 6).

2.3. Inflammatory gene expression

Neonatal rats (P4) were injected (i.p.) with saline or LPS (1 mg/kg) 1, 3, or 24h prior to being anesthetized (5% isoflurane, O₂ balance) and decerebrated. Medullas and cervical spinal cords were isolated in ice-cold artificial cerebrospinal fluid (aCSF; in mM: 120 NaCl, 26 NaHCO₃, 20 glucose, 2 MgSO₄, 1 CaCl₂, 3 KCl, and 1.25 Na₂HPO₄) and flash-frozen. Inflammatory gene expression (COX2, TNFα, and IL-1β) was analyzed on medullary and cervical spinal cord tissues stored at −80°C prior to sonication in Tri-Reagent (Sigma, St. Louis, MO, USA). Glycoblue reagent (Invitrogen, Carlsbad, CA, USA) was used to isolate total RNA. RNA (1 μg of total RNA) was reverse transcribed to complementary DNA (cDNA) using MMLV reverse transcriptase together with a cocktail of oligo dT and random primers (Promega, Madison, WI, USA), as previously described (Crain et al., 2009). qRT-PCR was performed with PowerSYBR green PCR master mix using an ABI 7500 Fast system. The relative expression of inflammatory genes in medullary and cervical spinal cord tissue homogenates was determined relative to 18s ribosomal RNA using the following primers:

• COX2: 5’-TGT TCC AAC CCA TGT CAA AA and 5’- CGT AGA ATC CAG TCC GGG TA;

• TNFα: 5’-TCC ATG GCC CAG ACC CTC ACA C and 5’-TCC GCT TGG TGG TTT GCT ACG;

• IL-1β: 5’- CTG CAG ATG CAA TGG AAA GA and 5’-TTG CTT CCA AGG CAG ACT TT;

• 18s: 5’- CGG GTF CTC TTA GCT GAG TGT CCC G and 5’- CTC GGG GCT TTG AAC AC.

Where possible, primers were designed to span introns (Primer 3 software) and were purchased from Integrated DNA Technologies (Coralville, IA, USA). Primer efficiency was assessed by use of standard curves, as previously reported (Crain et al., 2009). Gene transcripts were considered undetectable and not included in statistical analyses if their CT values fell outside of the linear range of the standard curve for that primer set (i.e. ≥34 cycles). The ddCT method was used to analyze qRT-PCR data (Livak & Schmittgen, 2001), as described previously (Crain et al., 2013; Nikodemova & Watters, 2011). Statistical analyses were performed on dCT values. Multivariate comparisons were performed using a two-way ANOVA (Sigma Stat version 11, Systat Software, San Jose, CA, USA) with a Holm-Sidak post hoc test. All data are expressed as mean fold change ± SD relative to the saline-injected pups.

2.4. Brainstem-spinal cord (BSSC) preparations

BSSCs from male and female neonatal rat pups were isolated as described previously (Greer et al., 1992; Suzue, 1984). In brief, neonates were anesthetized with isoflurane and decerebrated. The thoracic and cervical regions were isolated and placed in artificial cerebrospinal fluid (aCSF), containing the following (in mM): 120 NaCl, 3 KCl, 1.25 NaH₂PO₄, 1.0 CaCl₂, 2.0 MgSO₄, 26 NaHCO₃, 20 D-glucose, equilibrated with 95% O₂/5% CO₂. A dorsal laminectomy was performed to reveal the spinal cord, the lungs and heart removed, and a ventral laminectomy isolated the BSSC. Brainstems were transected at the pontomedullary junction. Isolated BSSC preparations were pinned ventral side up in a recording chamber (0.75–8.2 ml volume, 28 °C) with aCSF (continuously bubbled with 95% O₂/5% CO₂) and circulated by via a peristaltic pump (6–10 ml/min, MINIPULS3, Gilson, Inc., Middleton, WI). Fictive respiratory activity was recorded from the fourth or fifth cervical nerve rootlets using glass suction electrodes (internal diameter 70–100 μm). Recordings are amplified (x1000–10k), bandpass-filtered (0.1 Hz to 1kHz) (Model 1700 Differential AC amplifier, A-M Systems, Carlsburg, WA), integrated (τ = 50 ms) and rectified (LabChart, Version 7, ADInstruments; Axoscope, Molecular Devices). Preparations equilibrated for 40–50 minutes prior to data collection, after which bursts were averaged in 5-minute bins, and amplitude was normalized to baseline (first 5 minutes). Amplitude and frequency irregularity scores were calculated as previously described (Telgkamp et al., 2002), and were as follows: score of the nth cycle = 100*ABS[(P_n – P_n−1)/P_n−1] (ABS = absolute value, P_n = normalized amplitude, and P_n−1 = preceding normalized amplitude). Statistical significance was determined by two-way repeated measures ANOVA (5-minute bins) with post-hoc comparisons (Bonferroni test) using Graphpad Prism (Version 7, GraphPad Software, La Jolla, CA, USA) or Sigma Stat (Jandel Scientific Software, San Rafael, CA, USA). Differences were considered significant if p<0.05. Values are expressed as average ± SD. Poincaré plots for period were generated in R from periods obtained using Peak Analysis in Labchart (Version 7) to reveal burst pattern relationships by graphing the period between two bursts (T_n) versus the subsequent period (T_n+1). Statistical significance for rhythm disturbances after LPS was determined by individual Fisher’s exact tests (Graphpad Prism, Version 7, GraphPad Software, La Jolla, CA, USA).

2.5. Rhythmically active medullary slice preparations

Rhythmically active medullary slices, containing the preBötzinger Complex (preBötC, site of inspiratory rhythm generation, Rekling & Feldman, 1998; Smith et al., 1991), hypoglossal motor nucleus, and hypoglossal nerve roots, were isolated from neonatal Sprague-Dawley rats, as previously described (Ruangkittisakul et al., 2006). In brief, BSSCs (as described above) were pinned and thin slices (200 μm) cut using a vibratome (VT1000S, Leica, Nussloch, Germany) to visualize anatomical landmarks. Slices were compared to the neonatal rat brainstem atlas and a single 700 μm slice was taken at −0.35 to −0.45 mm from VIIc (Ruangkittisakul et al., 2006). Medullary slices were transferred to a recording chamber (8.2 ml volume, 28°C) with recirculated (10 ml/min) aCSF (continuously bubbled with 95% O₂/5% CO₂) via a peristaltic pump (MINIPULS3, Gilson, Inc., Middleton, WI). Extracellular potassium was elevated from 3 mM to 9 mM 30–60 minutes prior to the start of data collection to offset a loss of tonic excitatory inputs and prevent the gradual slowing of fictive respiratory activity in medullary slice preparations (Ruangkittisakul et al., 2006; Ruangkittisakul et al., 2007; Smith et al., 1991). Fictive respiratory activity was recorded from hypoglossal nerve rootlets in rhythmic slice preparations using glass suction electrodes (internal diameter 70–80 μm). Recordings were amplified (x10k), bandpass-filtered (300Hz to 1kHz) (Model 1700 Differential AC amplifier, A-M Systems, Carlsburg, WA), integrated (τ = 50 ms) and rectified (LabChart, ADInstruments). Bursts were averaged in 5-minute bins and amplitude normalized to baseline (first 5 minutes). Amplitude irregularity scores were calculated as described above. Statistical significance was determined by a two-way repeated measures ANOVA with post-hoc comparisons using the Bonferroni test. Differences were considered significant if p<0.05 (Graphpad Prism, Version 7, GraphPad Software, La Jolla, Ca USA). Values are expressed as averages ± SD. Poincaré plots were generated as described above.

3. Results

3.1. Neuroinflammation in the medulla and cervical spinal cord peaks 3h after LPS

Neonatal systemic LPS increased inflammatory gene expression for COX2, TNFα, and IL-1β genes in homogenate medullary tissue relative to saline controls (Fig. 1A). In the medulla after 1h LPS, COX2 (2 ± 0 fold change, p<0.001) and TNFα (13 ± 3 fold change, p<0.001) mRNA significantly increased compared to saline controls (Fig. 1A), suggesting systemic inflammation quickly induces neuroinflammation. In contrast, IL-1β mRNA was not changed 1h after LPS (2 ± 0 fold change, p=0.069). By 3h, expression of all three genes peaked (COX2: 15 ± 3 fold change, p<0.001; TNFα: 32 ± 8 fold change, p<0.001; IL-1β: 3 ± 1 fold change, p=0.008), and were increased relative to saline controls. Gene expression for all three genes remained elevated 24h post-LPS (COX2: 2 ± 0 fold change, p=0.001; TNFα: 10 ± 2 fold change, p<0.001; IL-1β: 3 ± 1 fold change, p=0.02) compared to saline, suggesting neuroinflammation persists at least 24h in the medulla following an inflammatory stimulus.

Figure 1.

Neonatal systemic inflammation (lipopolysaccharide, LPS, 1 mg/kg, i.p.) induces acute, CNS inflammation in respiratory control regions. In the medulla (A), COX2 and TNFα gene expression significantly increased 1h post-LPS (n = 6/group), with peak gene expression for all genes (COX2, TNFα, and IL-1β) at 3h (n = 6/group). At 24h post-LPS (n = 6/group), all three genes were significantly reduced from earlier time points, but remained elevated above saline controls. Note: the scale for the IL-1β gene expression is smaller than for COX2 and TNFα gene expression. After LPS in the cervical spinal cord (B), COX2 and TNFα gene expression increased 1h post-LPS, with peak expression at 3h of all three genes (n = 6/group). At 24h, all three genes were significantly reduced from 3h, but only TNFα gene expression remained elevated above saline controls. *p < 0.05, **p < 0.01, ***p < 0.001 significant difference within LPS and saline groups; ^#p < 0.05, ^##p < 0.01, ^###p < 0.001 significant difference between saline and LPS groups; two-way ANOVA with Holm-Sidak post-hoc.

In the cervical spinal cord (Fig. 1B), COX2 and TNFα mRNA increased at 1h post-LPS (COX2: 7 ± 1 fold change, p<0.001; TNFα: 8 ± 2 fold change, p<0.001), while IL-1β mRNA did not change (5 ± 2 fold change, p=0.056) relative to saline. At 3h, LPS increased COX2 (14 ± 4 fold change, p<0.001), TNFα (27 ± 9 fold change, p<0.001), and IL-1β (7 ± 4 fold change, p=0.04) mRNA compared to saline. However, by 24h post-LPS, TNFα mRNA remained elevated (9 ± 2 fold change, p<0.001), while COX2 (2 ± 0 fold change, p=0.285) and IL-1β (2 ± 1 fold change, p=0.835) mRNA returned to baseline. Similar to the medulla, COX2 and TNFα gene expression was upregulated after 1h with a peak at 3h, whereas IL-1β gene expression increased only at 3h in the spinal cord. No significant differences were evident between time points after saline, except an increase in IL-1β expression at 24h (2 ± 0 fold change) compared to 3h (1 ± 0 fold change, p=0.027).

3.2. Rhythm disturbances after systemic inflammation

Amplitude and frequency disturbances were observed following systemic inflammation. Rhythm disturbances after saline and all LPS doses were evident as large depolarizations at 1h (Table 1), suggesting these disturbances are not inflammation mediated. Frequency and amplitude rhythm disturbances, defined as abnormal respiratory behavior characterized by high frequency rhythm, abrupt amplitude changes, large depolarizations, and irregular period were also seen 3h following LPS (0.1, 1, and 10 mg/kg), but not after saline. Rhythm disturbances were significantly different between 1 and 3h after saline; however, rhythm disturbances were not different between 1 and 3 h after 1mg/kg or 10 mg/kg, likely due to more overall rhythm disturbances in both 1 and 10 mg/kg LPS. At 3h after LPS, four different types of activity were observed: 1) a high frequency rhythm (0.5–1.2 Hz, Fig. 2B, 16% of LPS preparations), not obviously coupled to the respiratory-related motor activity; 2) abrupt, irregular changes in motor burst amplitude (Fig. 2C; 26% of LPS preparations); 3) large depolarizations unrelated to the respiratory-related motor bursts (Fig. 2D; 16% of LPS preparations); and 4) irregular period between bursts, and burst failure (Fig. 2E, 10% of LPS preparations). The period of high frequency bursting lasted 12 ± 3 min (range = 7–24 min). These episodes had a sudden onset, with maximum frequency increase within 1–4 min, and a slow decrease in frequency before an abrupt stop. The peak frequency of high frequency bursts was 0.71 ± 0.7 Hz (range = 0.55–0.97). These high frequency rhythms were “superimposed” on the regular respiratory rhythm (Fig. 2). Resetting of the respiratory rhythm did not occur after high frequency rhythm.

Table 1.

Rhythm disturbances are evident in all groups at 1h, but increase in incidence after 3h LPS in brainstem-spinal cord preparations. At 3h, rhythm disturbances occurred only following LPS (0.1, 1, and 10 mg/kg). LPS (1 and 10 mg/kg) increased the incidence of rhythm disturbances at 3h relative to saline treated controls. *p < 0.05, **p < 0.01 significant difference compared to saline; ^#p < 0.05 significant difference between time points; Fisher’s exact test.

Injection time	Treatment	Number of preparations with rhythm disturbances
1h post-injection	Saline	6/14
	1 mg/kg LPS	7/11
	5 mg/kg LPS	2/5
	10 mg/kg LPS	2/6
3h post-injection	Saline	0/9
	0.1 mg/kg LPS	4/10
	1 mg/kg LPS	6/11*
	10 mg/kg LPS	7/11**

Figure 2.

Systemic neonatal inflammation causes rhythm disturbances in brainstem-spinal cord preparations 3h after LPS. Respiratory-related burst frequency and amplitude were regular in preparations from saline-injected pups (A). Rhythm disturbances (B-E), indicated by the black bars under the compressed neurograms, are highlighted in the expanded traces. LPS (0.1 mg/kg) induced high frequency (0.5–1.2 Hz) rhythms, unrelated to respiratory-related bursts (B, arrows in inset) in between respiratory-related bursts (B, asterisks in inset). LPS (1.0 mg/kg) increased respiratory burst amplitude abruptly (C) and induced large depolarizations (D, arrows in inset) overshadowing respiratory bursts, but otherwise did not disrupt the rhythm regularity. LPS (1.0 mg/kg) disrupted the rhythm and led to burst failure (E).

The incidence of rhythm disturbances (Table 1) increased with increasing LPS dose at 3h (0.1 mg/kg = 40% of experiments; 1.0 mg/kg = 55% of experiments; 10 mg/kg = 70% of experiments), suggesting more severe inflammation increases the incidence of inflammation-induced rhythm impairment at 3h. Although an increase in the number of rhythm disturbances was seen after 0.1 mg/kg LPS, only 1 (p=0.0141) and 10 (p=0.0047) mg/kg LPS showed a significant increase in the number of rhythm disturbances compared to saline controls (Fisher’s exact test).

3.3. Dose- and time-dependent impairment of respiratory activity in BSSC preparations after systemic inflammation

To investigate the time and dose-dependent effects of acute systemic inflammation on respiratory motor output, cervical roots C4 or C5 respiratory-related activity of BSSC preparations were recorded after LPS (0.1–10mg/kg) at 1h (Fig. 3A) and 3h (Fig. 3B). Representative neurograms after 1h LPS show decreased amplitude over 90 min after 5 and 10 mg/kg LPS (Fig 3A). Rhythm disturbances and decreased frequency were seen 3h after all LPS doses (Fig. 3B), suggesting time-dependent effects of inflammation on respiratory activity.

Figure 3.

Systemic neonatal inflammation impairs fictive respiratory amplitude at 1h, and impairs frequency and causes rhythm disturbances at 3h in brainstem-spinal cord preparations. Representative integrated neurograms of C4/C5 fictive respiratory activity 1h (A) and 3h (B) following saline (i.p.) or lipopolysaccharide (LPS, 0.1, 1, 5, and 10 mg/kg, i.p.). Dotted lines denote baseline. At 1h after 5 mg/kg LPS (n = 5), amplitude (C, left panel) decreased from baseline (white bars) at 90 min (grey bars). At 90 min (n = 14), amplitude irregularity (C, right panel) increased from baseline 1 h after saline, but was unchanged within and between LPS groups. Burst frequency (D, left panel) decreased from baseline over 90 min after saline (n = 14), LPS 1 mg/kg (n = 11), and LPS 5 mg/kg. Frequency irregularity (D, middle panel) increased at 90 min after 5 mg/kg LPS compared to 1 mg/kg LPS (n = 5). Delta burst frequency (D, right panel) was not different between groups. At 3h, amplitude decreased from baseline over 90 min following LPS (0.1, 1, and 10 mg/kg) (E, left panel). Amplitude irregularity (E, right panel) increased from baseline over 90 min after 0.1 mg/kg LPS (n = 10). At 3h, baseline burst frequency (F, left panel) decreased after 1 mg/kg LPS (baseline, n = 21). In all groups over 90 min, burst frequency decreased from baseline (saline, LPS 0.1, 1, and 10 mg/kg). Frequency irregularity (F, middle panel) was not different within or between groups. Delta burst frequency (F, right panel) was not different between groups. *p < 0.05, **p < 0.01, ***p < 0.001 significant difference from baseline; ^#p < 0.05 significant difference between saline and LPS groups; two-way ANOVA with Bonferroni post-hoc.

Group data at 1h post-injection show LPS did not impair baseline amplitude (Fig. 3C, left panel). Over 90 min, amplitude was unchanged from baseline in saline (0.95 ± 0.14 normalized to baseline, p>0.9), 1 mg/kg LPS (0.97 ± 0.15 of baseline, p>0.9), and 10 mg/kg LPS (0.85 ± 0.22 of baseline, p=0.0737) groups, but was decreased after 5 mg/kg LPS (0.75 ± 0.43 of baseline, p=0.0002). At 90 min, amplitude after 5 mg/kg LPS was significantly reduced compared to saline (p=0.0231) and 1 mg/kg LPS (p=0.0127), but not 10 mg/kg LPS (p>0.9). The significant decrease is influenced by cessation of respiratory activity over 90 minutes in one preparation. This variability likely reflects animal to animal sensitivity to LPS, where some animals have greater sensitivity to LPS and thus greater respiratory impairment.

At 1h post-injection, baseline amplitude irregularity (Fig. 3C, right panel) was not different between treatments (p>0.9). Amplitude irregularity significantly increased from baseline at 90 minutes after saline (baseline: 0.15 ± 0.03; 90 minutes: 0.19 ± 0.04, p=0.0205), but not after 1 mg/kg LPS (baseline: 0.15 ± 0.02; 90 min: 0.15 ± 0.03, p>0.9), 5 mg/kg LPS (baseline: 0.12 ± 0.02; 90 min: 0.16 ± 0.02, p>0.9), or 10 mg/kg LPS (baseline: 0.14 ± 0.01; 90 min: 0.17 ± 0.03, p>0.9), suggesting LPS does not impact burst regularity.

Baseline burst frequency at 1h post-injection (Fig. 3D, left panel) decreased over 90 min, but LPS did not alter this effect. Burst frequency significantly decreased from baseline at 90 minutes following saline (baseline: 12 ± 4 bursts/min; 90 min: 10 ± 5 bursts/min, p=0.0030), 1 mg/kg LPS (baseline, 11 ± 3 bursts/min; 90 min, 9 ± 4 bursts/min; p=0.0002), and 5 mg/kg LPS (baseline, 8 ± 3 bursts/min; 90 min, 5 ± 2 bursts/min; p=0.0009), but not after 10 mg/kg LPS (baseline, 10 ± 4 bursts/min; 90 min, 9 ± 6 bursts/min; p>0.9). No differences in frequency were evident between saline and LPS groups at 90 min, suggesting inflammation does not enhance the time-dependent decrease in frequency at 1h post-injection.

Baseline frequency irregularity (Fig. 3D, middle panel) was not different between treatments (p>0.2). At 90 min, frequency irregularity increased after 5 mg/kg (0.2 ± 0.19) compared to 1 mg/kg LPS (0.07 ± 0.08, p = 0305), but it was unchanged within groups (p>0.1). Delta burst frequency (Fig. 3D, right panel) was not different between groups (p>0.2).

In contrast, at 3h post-LPS (Fig. 3E, left panel), amplitude decreased from baseline at 90 minutes following 0.1 mg/kg LPS (0.75 ± 0.17 normalized to baseline, p<0.001), 1 mg/kg LPS (0.78 ± 0.21 normalized to baseline, p=0.002), and 10 mg/kg LPS (0.80 ± 0.13 normalized to baseline, p=0.005), but not after saline (0.90 ± 0.30 normalized to baseline, p=0.164). Despite these within group decreases in amplitude, no between group differences at baseline (p>0.9) or at 90 min (p>0.154) were evident. Since decreases in amplitude occurred only after LPS, this suggests inflammation impaired burst amplitude over time.

Baseline amplitude irregularity (Fig. 3E, right panel) 3h post-injection was not different between treatments (p>0.289). Amplitude became more irregular at 90 min following 0.1 mg/kg LPS (baseline: 0.08 ± 0.05; 90 min: 0.12 ± 0.05, p=0.033), but was unchanged following saline (baseline: 0.07 ± 0.03; 90 min: 0.08 ± 0.03, p=0.354) and LPS 1 mg/kg (baseline: 0.07 ± 0.03; 90 min: 0.10 ± 0.05, p=0.083) and 10 mg/kg (baseline: 0.10 ± 0.05; 90 min: 0.11 ± 0.05, p=0.974). No differences at 90 min between treatments (p>0.9) were evident, suggesting inflammation did not increase burst irregularity.

Baseline frequency (Fig. 3F, left panel) was decreased compared to saline following 1 mg/kg LPS (p=0.041), but not 0.1 mg/kg LPS (p=0.071) or 10 mg/kg LPS (p=0.388). Frequency decreased from baseline over 90 minutes following saline (baseline: 10 ± 2 bursts/min; 90 min: 8 ± 1 bursts/min, p<0.001), 0.1 mg/kg LPS (baseline: 7 ± 2 bursts/min; 90 min: 6 ± 1 bursts/min, p<0.001), 1 mg/kg LPS (baseline: 7 ± 3 bursts/min; 90 min: 6 ± 4 bursts/min, p<0.001), and 10 mg/kg LPS (baseline: 8 ± 2 bursts/min; 90 min: 6 ± 2 bursts/min, p<0.001), but was not different between treatments at 90 minutes (p>0.312). Thus, inflammation after 1 mg/kg LPS impairs baseline respiratory frequency.

Frequency irregularity (Fig. 3F, middle panel) was not different within (p=0.861) or between (p=0.171) groups at baseline (saline: 0.11 ± 0.07; 0.1 mg/kg LPS: 0.16 ± 0.13; 1 mg/kg LPS: 0.13 ± 0.07; 10 mg/kg LPS: 0.34 ± 0.51) or at 115 min (saline: 0.09 ± 0.01; 0.1 mg/kg LPS: 0.14 ± 0.02; 1 mg/kg LPS: 0.28 ± 0.12; 10 mg/kg LPS: 0.18 ± 0.02). Delta burst frequency (Fig. 3F, right panel) was not different between groups (p=0.533).

3.4. Pattern disturbances after systemic inflammation

To further assess dysfunction in respiratory activity after neonatal inflammation, Poincaré plots were constructed to visualize relationships between bursts. Poincaré plots (Fig. 4A) highlight increased burst variation and distinct burst patterns at 1h after LPS in the form of doublets, a burst pattern marked by shortened periods between bursts. In contrast, doublets were not seen 3h after LPS (Fig. 4B), suggesting early temporal modulation of burst pattern after neonatal inflammation. Further, 3h after 1 mg/kg LPS, but not 0.1 mg/kg or 10 mg/kg LPS, bursts with very long periods, indicating a slowed rhythm, were observed (Fig. 4B), suggesting a dose-dependent effect on burst period after inflammation.

Figure 4.

Systemic neonatal inflammation increases burst-to-burst variation 1h, but not 3h, after LPS in brainstem-spinal cord preparations. Poincaré plots of period (A) across 90 minutes highlight increased variation in burst period after 1h LPS (1, 5, and 10 mg/kg). Example burst patterns demonstrate singlet (solid circle outline) and doublet burst (dashed circle outline) behavior. At 3h (B), the rhythm was regular, singlet activity in all preparations. After 1 mg/kg LPS, rhythm slowed over time (dotted circle outline), highlighted in the expanded trace. Unlike after 1h LPS, doublets did not appear 3 h post-injection. Insets display burst pattern on a larger scale.

3.5. Severe inflammation impairs respiratory activity in isolated medullary slices

After establishing inflammation impaired amplitude and frequency in preparations with intact medullary and spinal respiratory circuitry, we sought to target the dysfunction to more reduced medullary circuits controlling breathing. Hypoglossal respiratory-related activity of rhythmic slice preparations after LPS (1 and 10 mg/kg) at 1h (Fig. 5A) and 3h (Fig. 5B) was recorded. Representative neurograms show a gradual increase in amplitude over 115 min at 1h after 10 mg/kg LPS (Fig 5A), with no effect at 3h (Fig. 5B). This increase in amplitude at 1h after 10 mg/kg LPS reflects delayed onset of activity in 40% of preparations. Onset of activity was determined to be delayed if frequency was <50% of the maximum frequency at baseline after similar equilibration periods. Burst onset was quantified for all groups, whereby no significant differences between groups were evident (Fisher’s exact test). At 1h after 10 mg/kg LPS in medullary slices burst onset was delayed in two preparations, which was not observed in any other group.

Figure 5.

Systemic inflammation after 10 mg/kg LPS delays onset of bursting and impairs frequency in rhythmic slice preparations at 1h and 3h. Representative integrated hypoglossal neurograms of fictive respiratory activity 1h (A) and 3h (B) after saline (i.p.) or LPS (1 or 10 mg/kg, i.p.). Dotted line denotes baseline. Onset of respiratory activity was delayed after LPS 10 mg/kg (1h). At 1h, amplitude (C, left panel) increased from baseline (white bars) over 115 min (grey bars) after 10 mg/kg LPS (n = 5) and compared to all other groups. Amplitude irregularity (C, right panel) at 1h increased from baseline over 115 min after saline (n = 6). At 1h, baseline burst frequency (D, left panel) decreased after 10 mg/kg LPS. Burst frequency decreased within group over 115 min from baseline after saline (1h) and 1 mg/kg LPS (1h, n = 5). At 115 min, baseline frequency irregularity (D, middle panel) was increased after 10 mg/kg LPS compared to saline and 1 mg/kg LPS. At 115 min, frequency irregularity increased after 1 mg/kg LPS, but decreased after 10 mg/kg LPS. Delta burst frequency (D, right panel) was greater after saline and 1 mg/kg LPS compared to 10 mg/kg LPS. At 3h, amplitude (E, left panel) and amplitude irregularity (E, right panel) were unchanged within and between groups. Baseline burst frequency (F, left panel) decreased after 10 mg/kg LPS (n = 6) compared to saline controls (n = 6). Frequency decreased over 115 min from baseline after saline and LPS 1 mg/kg (n = 6). Frequency irregularity (F, middle panel) was unchanged within and between groups. Delta burst frequency (F, right panel) was significantly greater after saline compared to LPS 10 mg/kg. *p < 0.05, **p < 0.01, ***p < 0.001 significant difference from baseline; ^#p < 0.05, ^###p < 0.001 significant difference between groups; two-way ANOVA with Bonferroni post-hoc.

In contrast to the more intact BSSC preparation at 1h, amplitude (Fig. 5C, left panel) remained stable after saline (1.08 ± 0.32 normalized to baseline, p>0.9) and 1 mg/kg LPS (1.29 ± 0.31 normalized to baseline, p>0.9), but significantly increased from baseline at 115 min after 10 mg/kg LPS (3.41 ± 2.57 normalized to baseline, p<0.0001). The increase in amplitude after 10 mg/kg LPS was significant compared to saline (p<0.0001) and 1 mg/kg LPS (p=0.0001), and reflects the delayed onset of activity in 40% of preparations.

Amplitude irregularity at baseline 1h post-injection (Fig. 5C, right panel) was not different after saline or LPS (p>0.9). Amplitude irregularity increased after saline from baseline (0.29 ± 0.24) at 115 min (0.66 ± 0.86, p=0.0113), but was unchanged within 1 mg/kg LPS (baseline: 0.20 ± 0.08; 115 min: 0.44 ± 0.32; p=0.83) and 10 mg/kg LPS (baseline: 0.36 ± 0.08; 115 min: 0.34 ± 0.23; p>0.9) groups.

At 1h post-injection, baseline frequency (Fig. 5D, left panel) was decreased after 10 mg/kg LPS (10 ± 6 bursts/min, p=0.0225), but not after 1 mg/kg LPS (15 ± 2 bursts/min, p>0.9), compared to saline (saline: 16 ± 4 bursts/min). Over time, burst frequency decreased within saline (baseline: 15 ± 4 bursts/min; 115 min: 12 ± 3 bursts/min, p<0.0001) and 1 mg/kg LPS (baseline: 15 ± 2 bursts/min; 115 min: 11 ± 3 bursts/min, p=0.0012), but not 10 mg/kg LPS (baseline: 10 ± 6 bursts/min; 115 min: 12 ± 3 bursts/min, p>0.9). Although frequency decreased after 1 mg/kg LPS, this decrease was not observed in groups after more severe inflammation (10 mg/kg LPS). Since frequency also decreased after saline, it is unlikely the drop in burst frequency is inflammation related. Despite these within group changes, frequency was not different between groups at 115 min (p>0.9). Similar to the changes in amplitude, systemic inflammation impairs baseline respiratory frequency at 1h after 10 mg/kg LPS.

Baseline frequency irregularity (Fig. 5D, middle panel) was decreased after saline (baseline: 0.09 ± 0.08; p=0.0066) and 1 mg/kg LPS (baseline: 0.04 ± 0.02; p<0.0001) compared to 10 mg/kg LPS (baseline: 0.19 ± 0.15). At 115 min, frequency irregularity increased after 1 mg/kg LPS (baseline: 0.04 ± 0.02; 115 min: 0.07 ± 0.03; p=0.0188), but decreased after 10 mg/kg LPS (baseline: 0.19 ± 0.15; 115 min: 0.11 ± 0.09; p=0.0007). Frequency irregularity at 115 min was not different between groups (p>0.2). Over 115 min, delta burst frequency (Fig. 5D, right panel) was significantly greater after saline (−4 ± 3 bursts/min, p=0.0035) and 1 mg/kg LPS (−4 ± 3 bursts/min, p=0.0084) compared to 10 mg/kg LPS (2 ± 5 bursts/min).

At 3h post-injection, inflammation did not impair baseline amplitude (Fig. 5E, left panel). Amplitude was unchanged from baseline at 115 min in all groups: saline (1.30 ± 0.55 normalized to baseline, n = 6, p=0.0942), 1 mg/kg LPS (1.05 ± 0.22 normalized to baseline, n = 6, p>0.9), and 10 mg/kg LPS (1.06 ± 0.33 normalized to baseline, n = 6, p>0.9), and was not different between groups at 115 min (p>0.6). Thus, acute systemic inflammation has no effect on amplitude in isolated medullary circuitry during the peak inflammatory response.

Amplitude irregularity at 3h post-injection (Fig. 5E, right panel) was not different between groups at baseline (saline: 0.21 ± 0.07; 1 mg/kg LPS: 0.18 ± 0.04; 10 mg/kg LPS: 0.21 ± 0.10, p>0.9) or at 115 min (saline: 0.20 ± 0.10; 1 mg/kg LPS: 0.19 ± 0.12; 10 mg/kg LPS: 0.25 ± 0.10, p>0.9). There were no differences within groups at 115 min (p>0.9). Thus, inflammation at 3h does not affect amplitude irregularity in rhythmic slices.

At 3h post-injection, baseline burst frequency (Fig. 5F, left panel) was decreased after 10 mg/kg LPS (13 ± 2 bursts/min, p=0.0229), but not after 1 mg/kg LPS (14 ± 2 bursts/min, p=0.2414), compared to saline controls (16 ± 3 bursts/min). Frequency decreased over time within saline (baseline: 16 ± 3 bursts/min; 115 min: 14 ± 3 bursts/min, p<0.0001) and 1 mg/kg LPS groups (baseline: 14 ± 2 bursts/min; 115 min: 12 ± 2 bursts/min, p=0.0008), but not 10 mg/kg LPS (baseline: 13 ± 2 bursts/min; 115 min: 13 ± 3 bursts/min, p>0.9). No differences were evident between groups at 115 min (p>0.5). While frequency decreases over time, inflammation does not significantly alter this decrease.

Frequency irregularity (Fig. 5F, middle panel) was not different within (p=0.7490) or between (p=0.9354) groups at baseline (saline: 0.06 ± 0.04; 1 mg/kg LPS: 0.07 ± 0.03; 10 mg/kg LPS: 0.08 ± 0.03) or at 115 min (saline: 0.07 ± 0.07; 1 mg/kg LPS: 0.06 ± 0.09; 10 mg/kg LPS: 0.06 ± 0.04). Delta burst frequency (Fig. 5F, right panel) at 115 min was significantly greater after saline (−3 ± 3 bursts/min) compared to 10 mg/kg LPS (0 ± 2 bursts/min, p=0.0216), but not 1 mg/kg LPS (−2 ± 3 bursts/min, p>0.9).

3.6. Doublets after systemic inflammation do not originate within medullary respiratory circuitry

Poincaré plots (Fig. 6) were used to visualize the relationship between consecutive bursts and identify the presence of doublets. Normal, single burst, rhythmic activity with regular periods were evident in all groups as clusters with a linear relationship. As the period (n) increases, the subsequent period (n+1) similarly increases. This is evident as burst clusters in a positive linear relationship from 0. Such rhythmic activity reflects the singlet rhythmic burst pattern. In contrast to BSSC preparations, doublets were not present following LPS (1 and 10 mg/kg) at 1h (Fig. 6A) or 3h (Fig. 6B). After 10 mg/kg LPS, bursts with increased period were evident as outlier bursts (Fig. 6A, right panel), which corresponds to the slow onset of burst activity. Thus, the emergence of doublets after inflammation is not driven by respiratory circuitry within medullary slices.

Figure 6.

Burst-to-burst variation increased 1h after 10 mg/kg LPS in rhythmic slice preparations, but was unaffected at 3h. At 1h post-injection, Poincaré plots of period over 115 min (A) show regular, rhythmic, singlet bursts (solid circle outline) in all groups; however, interburst interval increased after LPS 10 mg/kg group (dotted circle outline). At 3h post-injection, Poincaré plots of period (B) show regular, rhythmic, singlet bursts (solid circle outline) in all groups (saline, LPS 1 mg/kg, and LPS 10 mg/kg); however, interburst interval increased after LPS 10 mg/kg group (dotted circle outline) Expanded traces below highlight regularity. Insets display burst pattern on a larger scale.

4. Discussion

To understand the temporal and dose-dependency of systemic neonatal inflammation on respiratory activity, we investigated the effect of increasing LPS dose (0.1–10mg /kg) at 1h and 3h on fictive respiratory activity. Although neonatal infections can have deadly consequences (Heron, 2018) and neonatal respiratory impairment during infection is common (Campion et al., 2006; Fetter et al., 1995; Hofstetter et al., 2008; Ollikainen et al., 1993), the effect of neonatal inflammation on neural circuitry controlling breathing remains poorly understood (Camacho-Hernández et al., 2019; Lorea-Hernández et al., 2016; Lu et al., 2012). Further, we know little about the impact of timing and magnitude of early life inflammation on neonatal respiratory activity. Here, we demonstrate the timing of neonatal respiratory impairments after systemic inflammation correlates with the magnitude of the inflammatory response. We further demonstrate dose-dependent impairments of respiratory circuitry modulating pattern, where LPS impairs amplitude at 1h (5 mg/kg) and 3h (0.1,1, and 10 mg/kg), while 10 mg/kg LPS impairs frequency (3h), suggesting differential respiratory circuit sensitivities to inflammation. Sickness behaviors and physiological responses to infection in adult humans after systemic LPS are also time and dose-dependent (Grigoleit et al., 2011). In contrast to our results, increasing LPS dose (0.8 ng/kg) increased anxiety and lowered mood compared to a lower LPS dose (0.4 ng/kg) or saline, while the lower LPS dose increased heart rate earlier (2h after LPS) than the high dose (3h after LPS) (Grigoleit et al., 2011). Thus, the interactive effects of time and dose on neural networks controlling breathing are different than the interactive effects on sickness behavior, further supporting neural circuit-specific effects of inflammation. Establishing the interaction between time and dose on respiratory activity has implications for understanding neonatal vulnerability to respiratory dysfunction during neonatal infection.

While others reported brainstem inflammation in neonates (Balan et al., 2011; Jafri et al., 2013) and adults (Quan et al., 1999) challenged with systemic LPS, this study is the first to show CNS inflammation in both the neonatal medulla and cervical spinal cord, two regions important for respiratory control. While systemic LPS can induce brainstem inflammation as early as 2h (Balan et al., 2011; Jafri et al., 2013), our study shows LPS increases inflammatory gene expression (TNFα, COX2 and IL-1β) as early as 1h, with effects peaking at 3h, and returning to baseline by 24h in the spinal cord (COX2 and IL-1β). Thus, while different regions of the respiratory circuitry have similar overall temporal changes in inflammatory gene expression, each gene has distinct temporal dynamics. While COX2 and TNFα are both upregulated at 1h and 3h after LPS in the medulla and cervical spinal cord, IL-1β is not upregulated until 3h in the medulla and cervical spinal cord. These differences in temporal changes between genes may be due to gene-dependent differences in mRNA stability (Hao & Baltimore, 2009) or fluctuations due to circadian rhythms (Keller et al., 2009) contributing to temporal changes in inflammatory gene expression after systemic inflammation. Thus, we hypothesize the change in IL-1β mRNA after saline reflects natural changes in inflammatory signaling between different animals. Understanding the changing relationship between inflammatory genes will help identify key inflammatory signals leading to specific neonatal respiratory disruptions.

The temporal dynamics of inflammatory genes correlate with more severe impairments in respiratory activity. Early in the inflammatory response (1h), inflammatory impairments in respiratory frequency were observed in isolated medullary networks controlling breathing, suggesting impairment of preBötzinger complex (preBötC) circuits soon after the inflammatory stimulus. However, other brainstem networks modulating breathing demonstrated different temporal dependence. Decreased amplitude was observed at 1h in BSSC preparations, an effect absent at this time in isolated medullary slices. The impairments in respiratory activity were more severe at 3h post-LPS (during the peak inflammatory response), where amplitude and baseline frequency decreased, and rhythm disturbances increased in BSSC preparations. At 3h, irregular and abrupt increases in respiratory amplitude were prominent, accompanied by decreased frequency. Interestingly, inflammation led to greater impairment in medullary slices at 1h compared to 3h, reflected by delayed onset in activity and impaired baseline frequency. Similar to the 1h rhythmic slice preparations, burst frequency at 3h decreased over time in both saline and 1 mg/kg LPS rhythmic slice preparations, suggesting this decrease is not related to the inflammation and is likely due to loss of excitatory inputs and gradual run down of these preparations. However, unlike the more intact BSSC preparation, rhythm and pattern disturbances were not detected at 3h in rhythmic slices, suggesting regional differences in the temporal susceptibility to systemic inflammation on neural circuitry controlling breathing. Taken together, the early inflammatory response appears to impair frequency, while peak inflammatory response impairs amplitude, frequency, and burst pattern. These results suggest the preBötC is sensitive to low-levels of systemic inflammation. Since the peak inflammatory response impairs amplitude, frequency, and burst pattern, we hypothesize central inflammation more broadly impairs central respiratory circuitry, leading to greater overall impairment. The lack of similar impairments in rhythmic slices suggest this broad impairment is not solely due to preBötC impairment, but also impairs pattern forming circuits, pre-motoneurons, motoneurons, and neuromodulatory inputs. Future studies are needed to characterize the extent to which each aspect of the circuitry is impaired.

Neonatal inflammation also induced a unique burst pattern in the form of doublets 1h after LPS in BSSC preparations, but were rarely observed after saline in BSSC preparations or in rhythmic slices. While previous work suggests doublets originate within the preBötC (Kam et al., 2013; Li et al., 2016; Lieske et al., 2000; Ruangkittisakul et al., 2008), the few doublets observed 1h after LPS in rhythmic slices suggest inflammation impacts circuitry not contained within the rhythmic slice (e.g. beyond the preBötC), but contained within the BSSC. While the identity of the region mediating doublets is unclear, the pFRG can modulate burst pattern, since doublets appear during anoxia, unless BSSC preparations are transected between the facial motor nucleus and preBötC (Taccola et al., 2007). While Taccola et al. (2007) did not investigate doublet formation after inflammation, their results suggest medullary regions beyond the preBötC, such as the pFRG, induce doublet formation in response to stressful stimuli. Additionally, local disinhibition of the NTS through injection of a GABA_A antagonist disrupts respiratory patterns (Dhingra et al., 2019), suggesting inputs from the NTS are important in maintaining normal respiratory pattern. Thus, disruption of NTS-related inputs after systemic inflammation may contribute to the respiratory disturbances or doublets observed here.

Neonatal infections range in severity, but the link between inflammation severity and impaired respiratory circuitry has not been established. Here, we report a nonlinear, dose-dependent impairment of fictive respiratory activity in BSSC preparations. Lower LPS doses (1 and 5 mg/kg LPS) were associated with the greatest impairments in amplitude and frequency in BSSC preparations, while a higher dose (10 mg/kg LPS) was associated with frequency impairment in rhythmic slice preparations. The results from neonatal BSSC preparations are in contrast to in vivo work in adult rats, showing increased breathing frequency impairment with increasing intratracheal LPS dose (van Helden et al., 1997). Our results are also in contrast to worsening outcomes (sickness behaviors and pain sensitivity) in adult humans with increasing LPS dose (Wegner et al., 2014). However, these previous studies were conducted on adults in vivo (rats and humans), underscoring the importance of developmental age in understanding the influence of inflammation on breathing. While anatomical contributions to differences in respiratory-related activity were observed in this study, no differences in inflammatory gene expression were observed between anatomically distinct regions (medulla and spinal cord). However, only a subset of inflammatory markers were measured here, leaving the possibility that other inflammatory signals are region-specific and contribute to differences in respiratory activity after inflammation. Regional differences (medulla vs. spinal cord) in inflammatory gene expression occurred after chronic intermittent hypoxia (Smith et al., 2013), suggesting different inflammatory responses within these CNS regions are also stimuli specific (chronic hypoxia vs. LPS). This further indicates different regions of the CNS have different propensities to respond to inflammation. Further studies are needed to understand the origin of these differences, whether it is due to differences in microglia and astrocytes in different anatomical regions or different regional inflammatory signaling in the CNS.

Several other studies have investigated bath application of LPS on respiratory activity. Continuous bath application of LPS decreased respiratory activity in rhythmic slice preparations (Camacho-Hernández et al., 2019; Lorea-Hernández et al., 2016; Lu et al., 2012), although the effects on amplitude and frequency depended on the concentration. Similar to the results shown here, respiratory amplitude decreased after bath application of LPS (200 ng/ml LPS, Lorea-Hernández et al., 2016), while frequency decreased after 5 μg/ml bath-applied LPS (Lu et al., 2012), also demonstrating dose-dependent impairment in respiratory activity. Additionally, bath-applied LPS prior to intermittent anoxia attenuated intermittent anoxia-induced LTF (Camacho-Hernández et al., 2019). An important contrast between the previous studies and our study is the method of LPS exposure. Bath application of LPS is directly inducing CNS inflammation (Camacho-Hernández et al., 2019; Lorea-Hernández et al., 2016; Lu et al., 2012), while intraperitoneal injections of LPS induces peripheral inflammation and indirectly central inflammation. As LPS does not cross the blood-brain barrier (Banks & Robinson, 2010), the profile of the inflammatory response and associated respiratory impairment after direct application of the LPS to the CNS likely differs from the indirect effects of activating the peripheral inflammatory signaling. Bath applied LPS will directly activate microglia through TLR4 (Fernandez-Lizarbe et al., 2009), while systemic LPS will activate peripheral TLR4 to induce production of peripheral cytokines (Levy et al., 2009), trigger active transport of peripheral cytokines (Banks & Kastin, 1997), and activate endothelial cells (Dauphinee & Karsan, 2006; Zeuke et al., 2002) and vagal afferents (Laye et al., 1995).

The mechanisms underlying the deficiencies in fictive respiratory activity are unknown; however, other studies suggest a role for microglia. Activated microglia impair respiratory networks by undermining the hypoxic ventilatory response after intratracheal LPS (Ribeiro et al., 2017). Additionally, LPS-activated microglia decrease in vitro respiratory amplitude (Lorea-Hernández et al., 2016) and attenuate plasticity in the preBötC (Camacho-Hernández et al., 2019). Such studies demonstrate activated microglia have the ability to impair respiratory network activity, and thus, are prime candidates for future studies investigating the effects of systemic LPS on isolated respiratory circuitry. As amplitude irregularity only changed within the saline group in BSSC and rhythmic slice preparations, this change in amplitude irregularity over time is not inflammation mediated. Since respiratory activity decreases over time in BSSC preparations (Fong et al., 2008; Smith & Feldman, 1987) and was not further impacted by inflammation, we hypothesize this decrease is independent of LPS-induced inflammation.

The higher incidence of inflammation-induced rhythm disturbances in BSSC preparations compared to medullary slices are likely due to more intact respiratory circuitry within BSSC preparations. BSSC preparations contain the inspiratory rhythm generator (preBötC), premotor neurons projecting to the spinal cord, and respiratory-related motoneuron pools (hypoglossal and phrenic). The BSSC preparations also contain a greater number of modulatory neurons (e.g., serotonergic, noradrenergic, dopaminergic) projecting to respiratory-related neurons and releasing neuromodulators, such as substance P, thyrotropin-releasing hormone, and somatostatin. These neuromodulators exert powerful influences on respiratory activity (Doi & Ramirez, 2008). The response at each level of the circuitry (preBötC, premotor neurons projecting to the spinal cord, phrenic motoneurons, and modulatory neurons) may have an additive effect, resulting in dysfunction of the respiratory network. Alternatively, inflammation may have different effects on individual components of the respiratory circuitry, which is supported by the rhythmic slice data presented here. Although medullary slices contain many of these same types of neurons, medullary slices lack the rostral medulla, including the RTN/pFRG and spinal cord structures. We hypothesize the lack of inflammation-induced rhythm disturbances in medullary slices is the result of loss of modulatory neurons within the rostral medulla or the spinal cord. A more intact circuit contains more respiratory-related regions, including neurons in the RTN/pFRG, rVRG, and cervical spinal cord, as well as more glia (including astrocytes and microglia) involved in respiration. Thus, systemic inflammation can impair multiple regions of respiratory control, resulting in greater dysfunction of the overall network. Different regions may also be more sensitive to inflammation. Further, the pFRG is contained within the BSSC preparation, but not in the medullary slice preparation. In neonates, the pFRG plays a role in inspiratory respiratory rhythm generation and pattern formation (Mellen et al., 2003; Oku et al., 2007), and thus may contribute to pattern changes in BSSC preparations through changes in communication with the preBötC.

5. Conclusions

Infections in neonates are prevalent (Cailes et al., 2018) and increase neonatal apneas (Anas et al., 1982; Bruhn et al., 1977; Campion et al., 2006), emphasizing a need to understand how inflammation impacts respiratory control. This study reveals time and dose-dependent impairments of neonatal respiratory circuitry after systemic inflammation. Here, we show for the first time a temporal impairment of respiratory activity, suggesting the highest vulnerability of neonates at the early and peak inflammatory responses. Further, low and moderate inflammation led to the greatest respiratory impairments, suggesting even modest infections have the potential to significantly affect neonatal respiratory activity. Inflammation-induced respiratory impairment was most pronounced in more intact circuitry, indicating anatomical differences in sensitivity to inflammation. In conclusion, this study furthers our understanding of the complicated, dynamic effects of inflammation on neonatal respiratory circuitry.