The Monash Animal Research Platform supplied the pregnant ewes (Merino-Border Leicester cross) used in this study. The study was approved by the Monash Medical Centre Animal Ethics Committee A (2021/02) under guidelines established by the National Health and Medical Research Council of Australia Code of practice for the care and use of animals for scientific purposes.

Preterm lambs were delivered from 13 time-mated twin-bearing pregnant ewes via caesarean section at 126-7 days of gestational age. The pregnant ewes were administered betamethasone via an intramuscular injection (12 mg; Celestone, Merck Sharp & Dohme, Australia) 48 and 24 h before delivery to accelerate the maturation of the fetal lungs, as used in clinical practice for expected preterm birth (Roberts and Dalziel, 2006). Ewes were fasted for 16 h before surgery with ad-libitum access to water. General anaesthesia of the ewe was induced via injection of 5% sodium thiopentone (20 mg/kg; Pentothal; Boehringer Ingelheim Australia) into the jugular vein. The ewe was then intubated to allow for positive pressure ventilation (EV500 Anaesthesia Ventilator, Campbell, UCLO Engineering, NSW, Australia). Inhaled isoflurane (1.5%–2.5%) in O₂ was used to maintain general anaesthesia. Continuous monitoring of the ewe’s end-tidal carbon dioxide (CO₂) level, heart rate (HR) and oxygen saturation (SpO₂) was performed during the entire experimental preparation (Surgivet Advisor Vital Signs Monitor, Smiths Medical). Ventilatory settings and fraction of inspired oxygen were adjusted to maintain the ewe’s oxygen saturation (SpO₂) at 95%–100% and end-tidal CO₂ between 35–45 mmHg.

Surgery was performed under clean but not aseptic conditions as described previously (Inocencio et al., 2021). A maternal midline abdominal incision and hysterotomy were performed, via which the fetal head, neck and forelimbs were exteriorized for fetal instrumentation, whilst maintaining the placental-umbilical circulation. An incision was made in the left cubital fossa of the fetal forelimb to expose the median nerve. A silicon cuff with two multi-stranded copper wire electrodes 1 cm apart was placed around the nerve (Inocencio et al., 2021). The fetus was then intubated, lung liquid passively drained and exteriorised onto the maternal abdomen. Ventilation of the preterm lamb was commenced using warm humidified air and volume-guaranteed mode at ∼5 mL/kg (Babylog8000Plus; Draeger; Germany), whilst still maintaining the placental-umbilical circulation. After 1 min of ventilation, the preterm lamb received surfactant via the endotracheal tube (100 mg/kg Curosurf; ChiesiPharma; Italy). After 3 min of ventilation, the umbilical cord was clamped and the lamb was transferred to an infant resuscitaire (Fisher and Paykel Healthcare, East Tamaki, New Zealand). The lamb’s body temperature was monitored with a rectal probe and maintained at 39°C–40°C throughout the study. Inhaled isoflurane (1%–1.5%) was used to maintain anaesthesia of the lamb. A pulse oximeter probe (Radical 4, Masimo Frenchs Forest NSW, Australia) was placed on the lamb’s right forelimb to measure SpO₂. After delivery of both twin lambs, ewes were euthanised using intravenous pentobarbitone (100 mg/kg; pentobarbitone sodium 325 mg/mL, Virbac, Milperra, NSW, Australia).

Following delivery and transition to the infant resuscitaire, a single-lumen catheter (5Fr, 1270.05 polyurethane catheter, Vygon, Paris; France) was inserted retrogradely into one umbilical artery to allow continuous measurement of mean arterial blood pressure (MABP) and heart rate (HR) (DTX Plus Transducer; Becton Dickinson). Umbilical arterial blood samples were collected regularly for measurements of arterial pH, lactate, partial pressures of oxygen (PaO₂) and carbon dioxide (PaCO₂) (ABL30; Radiometer, Copenhagen, Denmark). A double-lumen catheter was inserted into the umbilical vein to administer maintenance fluid of 10% glucose (at 3 mL/kg/h) and to administer sildenafil infusion (in lambs allocated to the sildenafil group). Ventilatory settings and fractional inspired oxygen were adjusted to maintain SpO₂ at >90% and PaCO₂ at 45–55 mmHg.

Silver cup electrodes were placed on the shaved scalp of the preterm lamb to record the somatosensory evoked potential (SEP) in bilateral somatosensory cortices (C3-5 and C4-6), according to published coordinates for key landmarks on the sheep’s head (Cook et al., 1987). Two-channel EEG using differential AC amplifiers with bandpass filters set at 1–35 Hz (Grass Instrument Model 79D, Quincy; United States) was used. Depth of anaesthesia was monitored using continuous EEG recording, to avoid the burst-suppression EEG pattern which indicates deep anaesthesia.

The light emitters and detectors (optodes) of a two-channel Near infrared spectroscopy (NIRO-200, Hamamatsu Photonics, Japan) were placed on the scalp over the bilateral somatosensory cortices, adjacent to the EEG electrodes. For each channel, the emitter and detector were placed 4 cm apart. A light-proof cloth was placed over the optodes to prevent ambient light from interfering with the NIRS recording. The changes in oxygenated haemoglobin (∆oxyHb), deoxygenated haemoglobin (∆deoxyHb) and total haemoglobin (∆total Hb = ∆oxyHb + ∆deoxyHb) in both hemispheres were continuously recorded.

The electrodes implanted around the left median nerve were connected to an isolated constant-current electrical pulse generator (ISOLATOR-11 stimulator isolation unit; Axon Instruments Inc., Foster City, CA, United States) to deliver pre-programmed trains of 2 msec pulses (LabChart 7, ADInstruments Pty Ltd., Bella Vista, NSW, Australia) (Nakamura et al., 2017). The amplitude of the stimulus for each lamb was set to a minimum required to evoke a visible twitch of the left forelimb (∼2–5 mA). Registration of median nerve stimulation as somatosensory stimulation was confirmed by recording the SEP. An electrical pulse was delivered to the median nerve at 1 Hz every 250 ms for over ∼40–80 repetitions and EEG signal from the contralateral cortex was averaged across the 250 ms to produce a SEP response; an example is shown in Supplementary Figure S1. The mean ± SEM latency and amplitude of the first negative or positive component (i.e., N1, P1) in the SEP for all lambs was 62.76 ± 3.55 msec and 15.25 ± 1.57 µV respectively. This is consistent with the first component of the SEP in preterm infants of ∼32 weeks of post-menstrual age occurring approximately 60 msec after transcutaneous stimulation of the median nerve (Tombini et al., 2009).

All physiological signals (MABP, HR, EEG, ∆total Hb, ∆oxyHb, ∆deoxyHb) were recorded digitally at a sampling rate of 1000 Hz using a data acquisition system (Powerlab; ADInstruments, Castle Hill, NSW, Australia) and associated software (LabChart 7, ADInstruments). To elicit and record the cerebral haemodynamic response in each animal, the median nerve was stimulated with stimulus trains of short (1.8 s), medium (4.8 s), or long (7.8 s) durations, using 2 ms pulses at a repetition rate of 3.3 Hz (Nakamura et al., 2017). The beginning of each stimulus train triggered the recording of all physiological signals over a 60 s window, preceded by a 5 s period of pre-stimulus recording baseline to determine the relative magnitude of change of the NIRS signals, MABP and HR following median nerve stimulation. Fifteen to 20 repeats of each stimulus train were delivered to produce a signal-averaged output of all physiological signals (EEG, MABP, HR, and NIRS signals) using Scope software (ADInstruments, Australia). The three stimulus train durations (1.8, 4.8, and 7.8 s) were delivered in random order.

One twin from each pregnancy was randomly allocated to the control group and the other twin allocated to the intervention group. Lambs within the intervention group were randomly allocated to receive either intravenous infusion of sildenafil (Revatio, 10 mg/12.5 mL) or iNO administered via the endotracheal tube (INOMAX, Mallinckrodt). The twin lambs were delivered and studied consecutively (Supplementary Figure S2). To account for the potential effect of prolonged anaesthesia, lambs were randomly allocated to either experimental protocol A where the first twin would be allocated as a control, or protocol B where the first twin would be allocated to receiving the intervention (sildenafil or iNO) (Supplementary Figure S2).

Based on neonatal studies on the time required to reach therapeutic plasma levels (Cochius-den Otter et al., 2020), sildenafil infusion at 133 mcg/kg/h (the recommended clinical loading dose) occurred for 1 h before the median nerve stimulations. iNO administration was started at 10 ppm for 10 min, then increased to 20 ppm for 20 min before nerve stimulations, which corresponds to iNO dosage used clinically for the treatment of PPHN (Peliowski et al., 2012). Median nerve stimulation with stimulus train durations of 4.8 and 7.8 s were performed in random order during the sildenafil infusion/iNO inhalation. The 1.8 s stimulation was not conducted in the sildenafil and iNO groups as the mild nature of the short stimulation might not properly reflect the haemodynamic changes induced by the treatments. At the end of the experiment, lambs were euthanised by intravenous injection of pentobarbital at 100 mg/kg.

The MABP and NIRS data across the 60 s post-stimulus window of recording were averaged every 1 s and expressed as changes (∆) from the averaged 5 s pre-stimulus baseline. Analysis and classification of the NIRS response patterns in the contralateral cortex (right hemisphere) were based on ΔoxyHb, as it presented a more robust signal-to-noise ratio than ΔdeoxyHb as reported previously (Bortfeld et al., 2007; Nakamura et al., 2017). A ‘positive’ or ‘negative’ response was defined as an increase or decrease of oxyHb respectively in the contralateral hemisphere, of more than 2 standard deviations from the pre-stimulus baseline data values.

To assess the duration of the positive cerebral haemodynamic response, we calculated the difference between the times at which the ΔoxyHb reaches its peak amplitude and descends to 40% of the peak amplitude, denoted as T_p40 (sec) (Supplementary Figure S3).

Statistical analyses were performed using Graphpad Prism 8 (Graphpad Software, La Jolla; United States). Normality was assessed by the Shapiro-Wilk test. The lambs’ weight and, latency and amplitude of SEP were compared between the control, sildenafil and iNO groups using one way ANOVA. Physiological parameters and arterial blood gas measurements were compared between those at the start and the end of the experiment, and between those recorded before and during sildenafil/iNO treatment, using the Student´s paired t-test for parametric data or Wilcoxon signed-rank test for non-parametric data. Using the Chi-square test for trend or the Fisher’s Exact test, the proportions of different patterns of haemodynamic response measured by NIRS were compared within the control group between the three stimulation durations, and also compared between the control group and intervention groups (sildenafil or iNO), for 4.8 and 7.8 s stimulations respectively. Effects of the three durations of stimulation on the ΔMABP (peak amplitude and time to peak) and ΔoxyHb responses (peak/nadir amplitude, time to peak/nadir and area under the curve (AUC) in the control group were compared using one-way repeated measures ANOVA, with Tukey’s post hoc analysis for parametric data and Friedman test for non-parametric data; for the positive and negative cerebral hemodynamic responses respectively. Pearson correlation analysis was performed to assess relationship between the peak/nadir ΔoxyHb and ΔMABP responses during stimulations. The ΔMABP and ΔoxyHb responses were compared between the control, sildenafil and iNO groups using one way ANOVA, for 4.8 and 7.8 s stimulations respectively. All results are expressed as mean ± SEM, and p < 0.05 was considered significant.

There was no difference between the average weights of preterm lambs allocated to the control (n = 11), sildenafil (n = 6), and iNO (n = 8) groups (Table 1). The total duration for which lambs were subjected to mechanical ventilation post-delivery was ∼2 h. All physiological parameters remained stable for the duration of the experiment and remained within the normal range for preterm lambs under these experimental conditions (Polglase et al., 2005) (Table 1). The latency of SEP was significantly lower in both sildenafil and iNO groups compared to the controls (Table 1; p = 0.05 and 0.008 respectively)

Based on changes in the oxyHb signal in the contralateral cortex, Table 2 summarises the incidence of positive, negative or no responses, and Table 3 summarises the amplitude and timing of the ΔMABP and ΔoxyHb responses following the three different durations of stimulation.

The peak changes in MABP and peak/nadir changes in ∆oxyHb in the contralateral cortex following median nerve stimulations are shown in Table 3.

The peak amplitude of the ∆MABP was significantly greater following 4.8 s stimulation compared to 1.8 s (p < 0.05), in the lambs which showed positive changes in ΔoxyHb. The time taken to reach peak MABP did not differ between the stimulation durations (Table 3).

The amplitude of peak/nadir change in the ∆oxyHb was not different between different durations of stimulation for the positive or negative responses respectively. The time taken to reach the nadir ∆oxyHb was significantly greater following 7.8 s stimulations compared to 4.8 s (p < 0.05) for the negative responses. The time taken to reach the peak ∆oxyHb was not different (Table 3). The contralateral ΔoxyHb responses showed no correlation with the peak change in MABP following all durations of stimulation, suggesting that the blood pressure change is not the determinant of the cerebral haemodynamic response (Supplementary Figures S4A–S4C).

Arterial blood gases, MABP, HR, SEP, and NIRS data obtained in preterm lambs at the start and after 60 min of sildenafil treatment or after 30 min of iNO inhalation are summarised in Table 4. Following administration of sildenafil or iNO, there were no significant changes in the baseline physiological or cerebral haemodynamic parameters.

Overall, the proportions of positive responses were higher in both intervention groups compared to the control groups, and reached significance with the 4.8 s stimulation (p < 0.05, Table 2).

During sildenafil infusion, five of the six lambs (83%) demonstrated a consistently positive ∆oxyHb response after both 4.8 and 7.8 s stimulations (Table 2; Figure 2E). Only one lamb (Figure 2E; pink circles) demonstrated a consistently negative response.

During iNO inhalation, all eight lambs demonstrated a positive ∆oxyHb response after 4.8 s stimulation (Table 2; Figure 2I). Following 7.8 s stimulation, five out of the eight lambs (63%) retained a positive ∆oxyHb response and three had a negative ∆oxyHb response (Table 2; Figure 3I: negative response—red, white and yellow circles).

The T_p40, which assesses the temporal characteristics of the positive ΔoxyHb response was significantly higher in lambs treated with sildenafil compared to both control and iNO groups (p = 0.04 and 0.03 respectively), following 4.8 s stimulations (Table 3). There was no difference in T_p40 between the three groups in response to 7.8 s stimulations.

There were no differences in the amplitude of or time to reach peak ΔMABP and ΔoxyHb between the 4.8 and 7.8 s stimulations within the sildenafil or iNO group; or between the three groups of lambs (control, sildenafil, and iNO) at 4.8 and 7.8 s stimulations respectively (Table 3).

Consistent with the results of this study, we have previously reported a similar increase in incidence of negative functional responses following prolonged somatosensory stimulations in fetal lambs of the same gestational age (Nakamura et al., 2017). Negative cerebral responses have also been observed in neonatal rodents (Kozberg et al., 2013; Zehendner et al., 2013) and human neonates (Bartocci et al., 2001; Kusaka et al., 2004; Zimmermann et al., 2012). The negative response involved reductions in both oxyHb and total Hb, and likely arose from arterial vasoconstriction. In neonatal rats, hindpaw stimulations also produce a negative haemodynamic response with global cerebral vasoconstriction, but conversely produces a positive response with pial artery dilatation in the adults rats (Kozberg et al., 2013), suggesting the development of functional hyperaemia with increasing age. In addition, prolonged somatosensory stimulations drive a decline in EEG activity and reduce CBF in the neonatal mouse (Zehendner et al., 2013), suggesting the inability to sustain neural processing during prolonged stimulations in the immature brain. The decrease in CBF may also be due to redistribution and hence ‘haemodynamic steal’ of blood towards an adjacent cortical region of neuronal activity (Kozberg and Hillman, 2016; Suarez et al., 2021). In contrast, prolonged stimulation in adult mice was associated with stable neural processing and increased regional CBF. Taken together, the resulting negative functional responses suggests a potential for regional cerebral hypoxia to develop in the preterm brain when neuronal activation is prolonged.