Do direct nose-to-brain pathways underlie intranasal oxytocin-induced changes in regional cerebral blood flow in humans?

Do nose-to-brain pathways mediate the effects of peptides such as oxytocin (OT) on brain physiology when delivered intranasally? We addressed this question by contrasting two methods of intranasal administration (a standard nasal spray, and a nebuliser expected to improve OT deposition in nasal areas putatively involved in direct nose-to-brain transport) to intravenous administration in terms of effects on regional cerebral blood flow during two hours post-dosing. We demonstrate that OT-induced decreases in amygdala perfusion, a key hub of the OT central circuitry, are explained entirely by OT increases in systemic circulation following both intranasal and intravenous OT administration. Yet we also provide robust evidence confirming the validity of the intranasal route to target specific brain regions. Our work has important translational implications and demonstrates the need to carefully consider the method of administration in our efforts to engage specific central oxytocinergic targets for the treatment of neuropsychiatric disorders.

The oxytocin (OT) system has been a promising research area in translational neuroscience over the past decade (1,2). Robust evidence from studies in preclinical models has demonstrated the importance of the central OT system in the development (3) and regulation of complex social behaviours (4,5), the modulation of pain processing (6), feeding behaviour (7) and neuroinflammation after brain ischemia (8). Harnessing the central OT system has been identified as a potential strategy for the development of targeted pharmacological interventions to help to improve outcome in several conditions for which efficacious treatments do not currently exist (e.g autism spectrum disorder (9), schizophrenia (10), migraine (11), stroke (8), obesity (12), Prader-Willi (13)).
Human studies almost exclusively target the central OT system by administering synthetic OT using nasal sprays, despite a lack of understanding of the mechanisms underpinning its pharmacodynamic effects. OT is a hydrophilic cyclic nonapeptide and is unable to cross the blood-brain barrier (BBB) in significant amounts (14). When administered orally, OT is degraded in the gut. For these reasons, the intranasal administration of OT has been favoured under the assumption that once in the nasal cavity, OT can reach the brain directly, bypassing the BBB (15). Two main mechanisms have been suggested to underpin this putative direct nose-to-brain transport (16). The first mechanism postulates internalization of the peptide into olfactory or trigeminal neurons innervating the posterior and middle areas of the nasal cavity, followed by axonal transport and central exocytosis. However, this mechanism would be slow (17) and therefore unlikely to be responsible for the central and behavioural effects that we observe within 15-60 minutes (18) of administering intranasal OT in humans. The second mechanism postulates that the peptide reaches the cerebrospinal fluid (CSF) and brain parenchyma via passive diffusion through perineural clefts in the nasal epithelium, which provide a gap in the blood-brain barrier (19). While some animal work is consistent with the existence of the second mechanism (20), there is a lack of robust evidence to support the existence of nose-tobrain transport in humans (17,21).
The lack of clarity regarding the mechanisms mediating the effects of intranasal OT in humans and the inconsistent results in existing studies and clinical trials using intranasal sprays to deliver OT (22) have raised questions about the validity of the intranasal route to administer OT to the brain. While very small amounts of intranasally administered OT have been reported to reach the CSF (17), peripheral concentrations in the blood are also concomitantly increased to supraphysiologic levels. The increase in plasma OT levels unavoidably engages OT receptors expressed throughout the body, including the gastrointestinal tract, heart, and reproductive tract (17). These systemic effects may impact indirectly on brain function and behaviour and could underlie, at least partially, the observed effects after intranasal OT. It is also possible that the small amount of synthetic OT that crosses the BBB from systemic circulation (17) may be sufficient to induce functional effects in the brain, either by directly activating receptors in the brain or by stimulating OT autoreceptors on OT-synthesizing hypothalamic neurons to induce the release endogenous OT in a positive feedback loop (23). These mechanisms might explain why OT when administered peripherally (e.g. intravenous infusion) can still impact behaviour (24)(25)(26)(27).
Intranasal drug delivery allows for fast absorption into the peripheral circulation of small molecules, avoiding undesirable first-order hepatic and intestinal metabolism (28,29). However, this route comes with the disadvantage of poor and unreliable control of the amount of the drug absorbed (28,29). Therefore, to maximize the chance of achieving significant translational advances, we need to confirm whether nose-to-brain pathways can be used to target the central OT system and whether they offer any advantage in relation to alternative methods. Otherwise, trials using intranasal OT may just result in a waste of scarce resources and missed opportunities to gain insight about whether OT presents a valuable drug therapy in humans. Once the validity of the intranasal route to deliver synthetic OT to the brain is confirmed, a second step would require the optimisation of methods for nose-to-brain delivery of OT.
The absence of a selective radiolabelled OT ligand in humans makes it impossible to directly examine the central penetration and distribution of synthetic OT after intranasal administration. An alternative strategy is to quantify and compare whole brain functional effects after OT administration. We have previously demonstrated the sensitivity of arterial spin labelling (ASL) magnetic resonance imaging (MRI) in quantifying changes in brain's physiology after intranasal OT administration (18), as reflected in changes in regional cerebral blood flow (rCBF) at rest. Changes in rCBF provide a quantitative, noninvasive pharmacodynamic marker of the effects of acute doses of psychoactive drugs (30,31), with high spatial resolution and excellent temporal reproducibility (32). As a result of neurovascular coupling, changes in rCBF are likely to reflect changes in neuronal activity rather than vascular effects (33), and they capture relevant differential neurotransmitter activity of neurochemical systems (18,34).
In this study, we used ASL MRI to investigate and compare changes in rCBF over time that follow intranasal and intravenous OT administration. The use of an intravenous comparator can illuminate whether intranasal OT induced changes on brain perfusion in humans are associated with nose-to-brain pathways or result from concomitant increases in systemic OT circulation. Standard nasal sprays are the predominant method of intranasal OT administration in humans, yet they were not designed to maximise deposition in the olfactory and respiratory epithelia that is thought to mediate nose-to-brain transport (35). For this reason, alongside a standard nasal spray, we used a nasal administration method (PARI SINUS nebuliser) that combines the production of small size droplets with vibration to maximize deposition in upper and posterior regions of the nasal cavity where the direct nose-to-brain transport putatively occurs (36).
First, we reasoned that if intranasal administration represents a privileged route for the central delivery of OT, then intranasal-induced changes in rCBF in brain regions typically associated with the effects of OT in the brain (e.g. the amygdala) (37)(38)(39) should not be explained by increases in plasmatic OT achieved after OT intravenous infusion. Second, if posterior regions of the nasal cavity are involved in direct nose-to-brain transport, then using a device that can increase deposition in these areas may result in a more robust pattern of changes in rCBF when compared to OT administration with a standard nasal spray.

Global CBF and subjective state ratings
We observed a linear decrease over time in participants' global CBF level and levels of alertness and excitement (Main effect of time-interval); however, there was no significant main effect of treatment or time-interval x treatment interaction (Figs. S1 and S2 and Table S1). A significant decrease in global CBF over time is commonly observed over long sessions in the scanner (probably due to decreases in the level of alertness, blood pressure, heart-rate or a combination of these factors (40)) and supports the inclusion of global CBF as a nuisance variable in all of our analyses. We did not observe any significant correlation between changes in global CBF and ratings of alertness or excitement over time (Fig. S3).

Whole brain flexible factorial analysis: treatment, time-interval and treatment x timeinterval effects
We first computed a flexible factorial model to investigate changes in rCBF as a result of the main effects of treatment, time-interval and time-interval x treatment interaction. We did not observe a significant main effect of treatment, however we found a significant main effect of time-interval on rCBF in several clusters across the brain, likely reflecting decreases in alertness and attention. Importantly, we observed a significant treatment x time-interval interaction in three clusters. These clusters extended over a network of regions including: 1) the left superior and middle frontal gyri and the anterior cingulate gyrus; 2) the right occipital gyrus, cerebellum, lingual and fusiform gyri, calcarine cortex, cuneus and inferior temporal gyrus; 3) the left putamen, caudate nucleus, insula, amygdala, parahippocampal gyrus, rectus gyrus and medial orbitofrontal cortex ( Fig. 2 and Table 1).

Whole-brain characterization of the changes in rCBF associated with each method of administration for each time interval, using paired T-tests
As our study is the first in man to investigate the pharmacodynamics effects of synthetic oxytocin on resting rCBF over an extended period of time when administered with any of the three methods of administration using a double-blind placebo-controlled crossover design, we followed up the flexible factorial model with an exhaustive series of paired T-tests at each time interval to investigate the direction of potential OT-induced changes in rCBF specifically for each treatment route (compared to placebo). Overall, we observed significant changes in rCBF over two temporal intervals, reflecting early (15-43 mins) and late (75-104 mins) effects of OT, which we describe in detail for each method of administration below. Results are summarized in Fig. 3 and Tables 2, 3 and 4. Given the novel and exploratory nature of these analyses we report the  exact FWE-corrected P-values for significant clusters without further adjustment for the number   of paired t-tests using conservative Bonferroni correction, and denote in Tables 2-4  region and the insula. We also observed increases in rCBF at 24-32 min post-dosing in two clusters, one spanning the right superior/middle/inferior occipital gyri, the calcarine sulcus and the cuneus bilaterally, and the other one the right cerebellum, and at 75-83 min post-dosing in a cluster spanning the postcentral gyrus, the superior/middle/inferior occipital gyri, the superior parietal gyrus, the inferior/middle temporal gyri, the precuneus, the calcarine sulcus and the cuneus, all in the right hemisphere. Accounting for plasma OT AUC had no effect on the changes in rCBF observed following administration with PARI SINUS ( Table 4).

Investigation of the apparent overlap in rCBF decrease between standard nasal spray and
intravenous OT administration. Decreases in rCBF observed after standard nasal spray and intravenous OT administration at 24-32 and 87-95 minutes post-dosing overlap anatomically to a substantial extent (Fig. 4). We thus followed up with a direct comparison of standard nasal spray vs. intravenous OT administration using paired sample T-tests for these time-intervals; we did not observe any significant differences in rCBF between the two administration methods. Individual  (Table S3). Accounting for plasma OT AUC in the paired T-test for the standard nasal spray vs. placebo and the intravenous OT vs.
saline comparisons eliminated all significant decreases in rCBF that were observed for each of these administration methods. However, accounting for plasma OT AUC had no effect on the changes in rCBF uniquely observed in the standard nasal spray versus placebo comparison.
Comparison of pharmacokinetic profiles among treatments.
OT reached peak plasma concentration at the end of dosing when administered intravenously or via nasal spray, and by ~ 15min post-dosing when administered with the nebuliser (Fig. 6a).
Intravenous peak plasma OT concentrations (C max ) were significantly higher than either intranasal administration method, while C max did not differ between the nasal administration methods (

Heart rate and heart-rate variability: treatment, time-interval and treatment x timeinterval effects
There were no significant main effects of treatment or time-interval, and no significant treatment x time-interval effects on heart rate or on any time domain, frequency domain or non-linear measures of heart-rate variability (Table S5).

DISCUSSION
This is the first in man study to investigate the pharmacodynamics of synthetic oxytocin on resting rCBF over an extended period of time when administered intravenously, with a nebuliser or a standard nasal spray. We used arterial spin labelling MRI as a pharmacodynamically sensitive signature and a double-blind placebo-controlled crossover design to achieve two aims.
First, we wanted to understand whether intranasal OT-induced changes on brain physiology in humans reflect privileged nose-to-brain delivery or result from concomitant OT increases in peripheral circulation. We reasoned that if intranasal administration represents a privileged route for the central delivery of OT, then intranasally induced changes in rCBF in brain regions typically associated with the effects of OT in the brain would not be explained by concomitant increases in plasmatic OT achieved after OT intravenous administration. Second, we sought to test if a new device for nasal administration of OT, designed to achieve increased deposition in the posterior regions of the nasal cavity putatively involved in direct nose-to-brain transport, could maximize intranasal OT-induced changes in rCBF, resulting in a more robust pattern of changes in brain's physiology. Our study yielded three key findings, which we discuss below in turn.
Our first key finding was the observation of OT-induced decreases in rCBF in the left amygdala and the anterior cingulate cortex with both the intravenous and standard nasal spray administration methods at overlapping temporal intervals. These decreases in rCBF in both the left amygdala and anterior cingulate cortex correlated with nasal spray or intravenous-induced increases in OT plasma concentrations, and became non-significant when these concomitant changes in OT plasma concentration were added as a covariate in the model. At the same time, concomitant change in plasma OT concentration did not correlate with or account for any of the remaining changes in rCBF, when OT was administered intranasally either with a standard nasal spray or the nebuliser.
The suppression of amygdala's activity constitutes one of the most robust findings in animal studies and intranasal OT studies in men (37,(41)(42)(43). For instance, the dampening of the amygdala BOLD response to negative affective stimuli after intranasal OT administration has been consistently shown in several studies using task-based fMRI (37,44,45). Similarly, human BOLD fMRI studies have implicated intranasal OT-induced decreases in BOLD in the anterior cingulate cortex in the modulation of social cognition (46), emotion (47) or fear consolidation (48) effects. These suppressive effects on BOLD match our observation of decreases in rCBF in these areas at rest -which we suggest is likely to reflect decreases in local metabolic demands associated with decreasing neural activity at rest. We provide first evidence that the intravenous infusion of OT echoes the effects of a standard spray administration on brain's physiology within key neural circuits at rest. These effects are consistent with previous observations of improved repetitive behaviours and social cognition in ASD patients after intravenous administration of OT (24)(25)(26)(27) and provide a possible mechanism by which these therapeutical effects on behaviour may arise. Therefore, our findings challenge the current assumption that key effects of intranasal OT on brain function and behaviour are entirely derived by direct nose-to-brain transport.
With respect to changes in rCBF induced by the intravenous administration of OT, there may be three possible mediating mechanisms. First, it is possible that the direct peripheral effects of OT on OT receptors expressed in vegetative territories, such as the heart, may be an indirect source of changes in areas of the interoceptive/allostasis network in the human brain, where the dorsal amygdala and the pregenual anterior cingulate cortex assume the role of visceromotor hubs (49). However, as we did not observe any effects of OT (irrespective of administration method) on heart rate or heart rate variability, possible OT-induced changes in cardiac physiology cannot explain the decreases in rCBF we observed in this study. Second, it is possible that the small amounts of OT that cross the BBB (17) (or a metabolite of OT that remains functional) is sufficient to induce changes in rCBF in brain regions of high density of the OT receptor, such as the amygdala, but not in other regions where the lower availability of the receptor would require higher local concentrations of the ligand to produce measurable effects (50). This hypothesis is in line with two recent studies. The first study reported that the intravenous infusion of labelled synthetic OT increased synthetic OT levels in the CSF in primates (51). The second study showed that circulating OT can be transported into the brain by a receptor for advanced glycation end-products (RAGE) on brain capillary endothelial cells and that this receptor mediated-transport is critical for some behavioural actions of OT such as parenting and bonding (52). Third, related to the above, it is possible that the small amount of synthetic OT crossing the BBB is sufficient to engage OT autoreceptors on the OT-synthesising neurons in the hypothalamus, inducing the release of endogenous OT in the brain in a positive feedback loop mechanism (53). However, convincing evidence supporting this hypothesis remains elusive. In fact, a recent study in primates that administered labelled OT and examined whether concomitant increases in the concentration of OT in plasma and CSF reflected synthetic (labelled) or endogenous OT reported that both plasma and CSF increases were driven by increases in the concentration of the synthetic labelled OT (51).
Our second key finding was the observation of increases in rCBF following intranasal administration (using either a standard nasal spray or the nebulizer) which could not be explained by concomitant increases in plasma OT. Indeed, there were no significant increases in rCBF (even at a lower threshold) when OT was administered intravenously. This finding provides upto-date pharmacodynamic evidence consistent with the contribution of direct nose-to-brain pathways for these effects of intranasal OT in humans. Intranasal OT-induced increases in rCBF at rest are likely to reflect OT-induced increases in local energetic demand resulting from enhanced neural activity. These are compatible with at least some of the reported enhancing effects of intranasal OT on facial processing (54), empathy and mentalizing (55), salience attribution (56) and their neural underpinnings.
Our third key finding was that while the application of the same nominal dose of intranasal OT (40IU) with the standard nasal spray and the nebuliser resulted in identical pharmacokinetic profiles, the patterns of OT-induced changes in rCBF were markedly different across the two method of intranasal administration. Given the similarity in pharmacokinetic profiles, we hypothesize that the difference in the patterns of OT-induced rCBF changes achieved with each method can only be explained by differences in the deposition of OT in the olfactory and respiratory regions and the parasinusal cavities which receive innervation from the olfactory and trigeminal nerves and may thus constitute important points of entry to the brain. It is possible that, as expected, the nebulizer achieved higher OT deposition in these areas (35) and hence resulted in increased amounts of OT reaching the brain (compared to the standard nasal spray).
Consistent with this hypothesis, we found that when administered with the nebuliser, OT robustly The fact that the nebuliser resulted in a different pattern of rCBF changes, with null or minimal overlap with the changes observed after the standard spray, instead of simply observing changes in the magnitude of the effects within the same areas, is surprising to some extent. Our initial hypothesis was that the nebuliser would result in a more robust pattern of changes that would be comparable to the spray but of higher magnitude and eventually include areas that could not be targeted neither by the spray nor by the intravenous administrations. However, we should acknowledge that this prediction would have mostly been valid if the pharmacodynamics of the rCBF response to OT followed a linear model -which does not seem to be the case at least for some brain areas. The few studies that have inspected the dose-response effects of intranasal OT on the BOLD response in the amygdala support an inverted-U shape curve of response by showing that deviating from an "optimal" dose may in fact result in lower or null effects (38,58).
We believe the complexity of the central OT signalling machinery (59) should be considered to interpret these findings. The OTR has been described to recruit different intracellular G protein (G s or G i ) pathways, depending on ligand, receptor and G protein type distribution and abundance (50,59). G s and G i activation typically result in opposite effects in terms of cellular function (60), meaning that in areas of high density of G i proteins higher amounts of OT may in fact result on inhibition of neuronal activity or null effects (50,59). This complexity might explain, for instance, why we did not observe changes in rCBF with the nebuliser in regions where the standard nasal spray produced effects, at our predefined statistical thresholds for this analysis.
Until a ligand allowing for direct quantification of in vivo penetration of OT in the brain after intranasal administration might be produced, a dose-response study using the nebuliser may allow us to gain further indirect insights about whether using the nebuliser may confer certain advantages regarding targeting the central OT system.
While our findings are consistent with the idea that direct nose-to-brain pathways could explain some of the changes in rCBF induced by intranasal OT, our study cannot provide evidence regarding the precise mechanisms underlying these effects. We believe that most evidence to date concurs on the idea that OT, when administered intranasally, may diffuse from increases in this area after the intravenous administration of the tracer were almost negligible (67). The increases of the concentration of the tracer in the olfactory bulb following its intranasal administration could be observed as soon as 30 minutes post-dosing, which fits the time-frame of the effects we report in our study. While our findings cannot illuminate the precise pathway through which intranasal OT may reach the brain, the fact that we observe distinct patterns of changes in rCBF with two different intranasal methods suggests that the changes we see in the brain are unlikely to be explained by local effects of intranasal OT in the nasal cavity. Expression of the OT receptor has been reported in human taste buds (68) and in the rat olfactory epithelium (69). Direct actions of OT on olfactory and trigeminal nerve bundles could parsimoniously explain the discrepancies between the changes observed after intranasal and intravenous administration of OT. If direct modulation of activity in these nerve bundles would account for all the changes we report for our intranasal administration methods, then one would expect that the changes observed in rCBF after intranasal administration of synthetic OT would be mostly restricted to the olfactory/trigeminal pathways and respective connected areas, which is not entirely supported by our data (for example, we have little evidence to believe that modulation of activity in olfactory or trigeminal nerves would result in changes in rCBF in the frontal gyrus). In fact, the effects we observed after synthetic OT map to areas where expression of receptors for OT (either the OTR or the AVPR1A) seem to be present, as per our current understanding of the distribution of these receptors in the human brain (70,71). While we cannot exclude that some of the effects we observed after intranasal OT may relate to direct modulation of activity in nerves bundles, this mechanism is unlikely to fully account for all the effects we report herein.
From a translational perspective, our findings emphasize the inadequacy of a one-fits-all approach in the administration of synthetic OT to target the central OT system in humans for the treatment of brain's disorders. Our findings indicate that some specificity may be achieved depending on the route used to deliver OT. Given that enhacement of brain's metabolism in areas such as the frontal gyrus, insula or occipital cortices may be restricted to the intranasal route, clinical applications aiming to target these circuits should thus prefer this route. An example could be, for instance, autism spectrum disorder, where the insula has been consistenly identified as a locus of hypoactivity (72). However, we should not completely discard the potential utility of the peripheral route -specially if the desired effect is to specifically decrease amygdala and anterior cingulate's metabolic activity in a targeted way, with minimal effects on other brain areas. This may be the case, for instance, of mood and anxiety disorders , where heightened amygdala/fear systems response has been consistently described (73,74). Systems of controlled sustained drug release for the peripheral circulation (i.e. transdermal controlled release (75)) already in place may provide an excellent opportunity to explore the clinical value of this route during chronic administrations in patients.
Our study faces certain limitations that we would like to acknowledge. First, the amount of OT administered intravenously was not chosen to mimic exactly the plasmatic concentrations achieved after intranasal administration, which would require the intravenous administration of a dose about 5 times lower (2IU) (58). Instead, we adopted a proof-of-concept approach, aiming to achieve consistently higher plasmatic concentrations of OT during the full period of scanning, eliminating the hypothesis that negative findings could be ascribed to insufficient dosing. Our approach increases our confidence in our interpretation that the unique changes in rCBF observed following intranasal administration cannot be explained by concomitant increases in plasmatic OT concentrations. Future studies should include an intravenous comparator that achieves pharmacokinetic profiles that are similar to those achieved with the intranasal methods. Second, our findings cannot be readily extrapolated to women, given the known sexual dimorphism of the OT system in the brain and behavioural responses to OT in humans (76)(77)(78). Third, although we tested for the potential effects of synthetic OT on cardiac physiology as a confounder, we acknowledge that potential effects on other peripheral systems need also be considered in the future, e.g. the reproductive or gastrointestinal tracts. Future studies need to compare the intranasal and intravenous administration of OT with the parallel administration of a specific nonbrain penetrant OT receptor antagonist to clarify the potential contribution of OT's signalling in the periphery to its effects on brain and behaviour. Fourth, we relied on a biomarker of brain physiology (rCBF) to probe the effects of different methods of administration of OT on brain function. This method exploits the well-established phenomenon of neuro-vascular coupling as a means of obtaining an indirect but sensitive signature of regional neuronal activity (79,80), but it is theoretically susceptible to potential confounding vascular effects (81). administration. Finally, the univariate analyses we present in this paper should not be used to define the temporal dynamics of the effects of synthetic OT on rCBF (as has been previously explored using pattern recognition analyses) (18). Univariate analyses allow us to localise rCBF effects in contrast to pattern recognition analyses where each voxel makes a weighted contribution to classification.
In conclusion, we provide the first robust physiological evidence supporting the existence of direct nose-to-brain pathways in humans, while also demonstrating that some of the key effects of synthetic OT in the human brain, when delivered by standard nasal sprays, can be explained by concomitant increases in peripheral OT levels post-dosing. Our results emphasize the inadequacy of a one-fits-all approach in the administration of synthetic OT to modulate brain function for the treatment of psychiatric or neurological conditions in humans, while highlighting the importance of optimizing the delivery of peptides to the brain through nose-to-brain pathways.

Participants
We approved the study. We determined sample size based on our previous validation study demonstrating that N=16 per group was sufficient to quantify standard nasal spray OT-induced changes in rCBF in a between-subjects design (18,85).

Study design
We employed a double-blind, placebo-controlled, triple-dummy, crossover design. Participants

Intranasal OT administration
For the intranasal administrations, participants self-administered a nominal dose of 40 IU OT (Syntocinon; 40IU/ml; Novartis, Basel, Switzerland), one of the highest clinically applicable doses (86). We have shown that 40IU delivered with a standard nasal spray induce robust rCBF changes in the human brain in a between-subjects, single-blind design study (18). For the intranasal administration, we used specially manufactured placebo that contained the same excipients as Syntocinon except for oxytocin. droplet diameter is roughly one tenth of a nasal spray and its mass is only a thousandth. The efficacy of this system was first shown in a scintigraphy study by (87). Since the entrance of the sinuses is located near the olfactory region, an improved delivery to the olfactory region was expected compared to nasal sprays. Other studies (35) have shown up to 9.0% (±1.9%) of the total administered dose to be delivered to the olfactory region, 15.7 (±2.4%) to the upper nose.

Intravenous OT administration
For the intravenous administration, we delivered 10 IU OT (Syntocinon injection formulation, 10IU/ml, Alliance, UK) or saline via slow infusion over 10 minutes (1IU/min). A 50-ml syringe was loaded with either 32 ml of 0.9% sodium chloride (placebo) or 30 ml of 0.9% sodium chloride with 2 ml of Syntocinon (10IU/ml). A Graseby pump was used to administer 16 ml of the compound (hence 10 IU of OT in total) over 10 minutes, at a rate of 96 ml/h. The ECG was monitored during the intravenous administration interval. We selected the intravenous dose and rate of administration to assure high plasmatic concentrations of OT throughout the observation period while restricting cardiovascular effects to tolerable and safe limits. A rate of 1IU per minute is typically used in caesarean sections and is considered to have minimised side effects (88,89).

Procedure
Each experimental session began with the treatment administration protocol that lasted about 22 minutes in total (Fig. 1). After drug administration, participants were guided to the MRI scanner, where eight pulsed continuous arterial spin labelling scans (each lasting approx. eight minutes) where acquired, spanning 15-104 minutes post-dosing, as detailed in Fig. 1. Participants were instructed to lie still and maintain their gaze on a centrally placed fixation cross during scanning.

Blood sampling and plasmatic OT quantification
We collected plasma samples at baseline and at five time-points post-dosing (as detailed in Fig.   1) to measure changes in the concentration of OT. Plasmatic OT was assayed by radioimmunoassay (RIAgnosis, Munich, Germany) after extraction, currently the gold-standard technique for OT quantifications in peripheral fluids (90). Details of the protocol for sample processing and radioimmunoassay quantification of plasmatic OT can be found elsewhere (91).

MRI data acquisition
We used a 3D pseudo-continuous Arterial Spin Labelling (3D-pCASL) sequence to measure changes in regional Cerebral Blood Flow (rCBF) over 15-120 min post-dosing. Labelling of arterial blood was achieved with a 1525ms train of Hanning shaped RF pulses in the presence of a net magnetic field gradient along the flow direction (the z-axis of the magnet). After a postlabelling delay of 2025ms, a whole brain volume was read using a 3D inter-leaved "stack-ofspirals" Fast Spin Echo readout (92) cerebral signal from the CBF map, by multiplication of the "brain only" binary mask obtained in step [2], with each co-registered CBF map; (4) normalization of the subject's T1 image and the skull-stripped CBF maps to the MNI152 space using the normalisation parameters obtained in step [2]. Finally, we spatially smoothed each normalized CBF map using an 8-mm Gaussian smoothing kernel. All of these steps were implemented using the ASAP (Automatic Software for ASL processing) toolbox (version 2.0) (94). The resulting smoothed CBF maps from each C-L pair were then averaged, using the fslmaths command implemented in the FMRIB Software Library (FSL) software applications (http://www.fmrib.ox.a.c.uk/fsl), to obtain a single averaged CBF map for each of the time-intervals depicted in Fig. 1.

Physiological data acquisition and processing
Heart rate was continuously monitored during the scanning period using MRI-compatible finger pulse oximetry while the participant rested in supine position, breathing spontaneously in the scanner. The data were recorded digitally as physiologic waveforms at a sampling rate of 50 Hz. Heart beats were firstly automatically detected using an in-house script and then visually inspected and manually cleaned for misidentified beats. Inter-beat interval values were then calculated. The resulting cleaned data were then transferred to Kubios HRV analysis software (MATLAB, version 2 beta, Kuopio, Finland) and a set of time domain (heart-rate (HR) and the root mean square of the successive differences (RMSSD), frequency domain (low (LF) and high (HF) frequencies spectral powers and high/low frequency spectral power ratio (HF/LF)) and nonlinear (Approximate entropy (ApEn), the SD1 and SD2 lines from the Poincare Plot and the detrended fluctuation scaling exponents DFAα1 and DFAα2) analysis measures were calculated.
A detailed description of the analysis methods used to calculate these measures have been described elsewhere (95,96). We decided to examinate a wide-range of different heart variability measures because previous studies have diverged in the metrics where they found effects of OT on heart rate variability (95). For instance, there is currently debate whether time and frequency analysis measures can be sufficiently sensitive to capture important (nonlinear) changes in heart rate time series, including those changes associated with OT administration (96)(97)(98). In addition to the manual cleaning of the data, we also employed a threshold-based method of artefact correction, as provided by Kubios, where artefacts and ectopic beats were simply corrected by comparing every RR interval value against a local average interval. The threshold value used was 0.35 seconds. Following current recommendations for heart rate data processing and analysis, if more than 5% of the beats required correction, then we decided to exclude these periods of observation (99). All of these metrics were calculated based on the data recorded for the whole duration of each scan (8 min). In some rare cases where artefacts could not be corrected, we only included the data if at least 5 min of acquisition free of artefacts could be analysed. We based our decision on the fact that at least 5 min of observation are required for pulse plethysmography to reflect heart rate variability as assessed by electrocardiography (100). The percentages of the total amount of available data used for this analysis after data quality control can be found in Supplementary Table S4.

Global CBF Measures
We extracted mean global CBF values within an explicit binary mask for grey-matter (derived from a standard T1-based probabilistic map of grey matter distribution by thresholding all voxels with a probability >.20) using the fslmeants command implemented in the FSL software suite.
We tested for the main effects of treatment and time-interval and for the interaction between both factors on global CBF signal in a repeated measures analysis of variance Treatment and Time as factors, implemented in SPSS 24 (http:// www-01.ibm.com/software/uk/analytics/spss/), using the Greenhouse-Geisser correction against violations of sphericity.

Subjective ratings
For the two subjective ratings of alertness and excitement collected at the three time points postdosing, we initially tested for the main effects of treatment and time interval and for the interaction between both factors, as previously described for global CBF. Second, we investigated the association between changes in global CBF signal and self-ratings of alertness and excitement over time, using within-group pooled correlation coefficients. For equal variances of the correlated variables in the four subgroups, pooled within-group correlation coefficients represent a weighted mean of the within-group correlation coefficients, weighted by the number of observations in each subgroup (101). They correspond to the result of statistically eliminating subgroup differences from the total group correlation coefficient. Since we only collected ratings at three time points, for this analysis we selected the global values of the scans that were closer in time to the moment of the ratings acquisition. We firstly inspected the equality of the covariance matrixes for each rating scale across the four treatment groups to decide whether to pool the four groups or not, using the mconvert command at SPSS. We then calculated the association between each rating scale and global CBF for each participant and averaged the covariance matrices to estimate the pooled within-group Pearson correlation coefficients. Results are reported at a level of significance α = 0.05.

Whole-Brain Univariate Analyses
We firstly implemented an analysis of covariance design, controlling for global effects on CBF, Our study is novel in multiple ways: it is the first study to investigate the pharmacodynamics effects of synthetic oxytocin on resting brain physiology over an extended period of time when administered with any of the three methods of administration we used in a double-blind placebocontrolled crossover design. It is also the first in man study regarding the effects of OT administered intravenously or with a nebuliser on rCBF. Since our study maps uncharted territory, we followed up the flexible factorial model with an exhaustive series of paired T-tests at each time interval to investigate the direction of potential OT-induced changes in rCBF specifically for each treatment route (compared to placebo).
We conducted whole-brain cluster-level inference for all analyses, reporting clusters significant at α = 0.05 using familywise error (FWE) correction and a cluster-forming threshold of P = 0.005 (uncorrected). Our statistical thresholds were determined a priori based on our own previous work investigating the effects of intranasal spray OT on rCBF in humans (18)

Association between OT-induced changes in rCBF and plasma OT concentration
To investigate if concomitant increases in peripheral OT were related to treatment-induced changes in rCBF, we extracted data from each significant rCBF cluster (adjusting for each treatment comparison contrast) in the paired sample T-tests and calculated Spearman correlation coefficients between these contrasts estimates and the AUCs reflecting individual differences in treatment-induced OT plasma concentrations for the corresponding method of administration. In this specific case, we employed the Spearman correlation coefficient because the number of observations used to estimate the correlation is small, which does not allow for an accurate verification of all implicit parametric analysis assumptions.

OT effects on cardiac physiology
Previous human studies have shown the ability of OT to affect cardiac physiology. Specifically, these studies have suggested that intranasal OT increases HRV at rest (108). HRV is an important index for the heart-brain interaction (109). Changes in HRV are accompanied by changes in the activity of several areas of the brain, including the amygdala (one of the areas of the brain most commonly implicated in OT effects on brain function and behaviour) (110,111). Thus, OTinduced changes in HRV, if existent, could account for, at least, some of the OT-induced changes in rCBF we identify herein. We compared mean HR, RMSSD, HF, LF, HF/LF ratio, ApEN, SD1, SD2, DFAα1 and DFAα2 between methods of administration to examine the extent to which OT administration induced changes in HR or heart rate variability, by using a repeated measures twoway analysis of variance. We used treatment and time interval as factors. We determined main effects of time interval and treatment, as well as their interaction, and used the Greenhouse-Geisser correction against violations of sphericity. We contained the family-wise error (FWE) rate at α =0.05 using the Benjamini-Hochberg procedure, which is a more powerful version of the Bonferroni adjustment that allows non-independence between statistical tests (112). Original p values (two-tailed) are reported alongside with values obtained after accounting for FWE.
All the analyses were conducted with the researcher unblinded regarding treatment condition.
Since we used a priori and commonly accepted statistical thresholds and report all observed results at these thresholds -the risk of bias in our analyses is therefore minimal, if not null.
Table S1 -Effects of treatment, time-interval and treatment x time-interval on global CBF and subjective ratings.        Comparison of absolute plasmatic bioavailability between spray and nebulizer. Data are presented as mean ± 1SEM. Statistical significance was set to p<0.05.