Effects of paracetamol (acetaminophen) on gene expression and permeability properties of the rat placenta and fetal brain

Background: Paracetamol (acetaminophen) is widely used in pregnancy and generally regarded as “safe” by regulatory authorities. Methods: Clinically relevant doses of paracetamol were administered intraperitoneally to pregnant rats twice daily from embryonic day E15 to 19 (chronic) or as a single dose at E19 (acute). Control samples were from un-treated age-matched animals. At E19, rats were anaesthetised, administered a final paracetamol dose, uteruses were opened and fetuses exposed for sample collection. For RNA sequencing, placentas and fetal brains were removed and flash frozen. Fetal and maternal plasma and cerebrospinal fluid were assayed for α-fetoprotein and interleukin 1β (IL1β). Brains were fixed and examined (immunohistochemistry) for plasma protein distribution. Placental permeability to a small molecule ( 14C-sucrose) was tested by injection into either mother or individual fetuses; fetal and maternal blood was sampled at regular intervals to 90 minutes. Results: RNA sequencing revealed a large number of genes up- or down-regulated in placentas from acutely or chronically treated animals compared to controls. Most notable was down-regulation of three acute phase plasma proteins (α-fetoprotein, transferrin, transthyretin) in acute and especially chronic experiments and marked up-regulation of immune-related genes, particularly cytokines, again especially in chronically treated dams. IL1β increased in plasma of most fetuses from treated dams but to variable levels and no IL1β was detectable in plasma of control fetuses or any of the dams. Increased placental permeability appeared to be only from fetus to mother for both 14C-sucrose and α-fetoprotein, but not in the reverse direction. In the fetal brain, gene regulatory changes were less prominent than in the placenta of treated fetuses and did not involve inflammatory-related genes; there was no evidence of increased blood-brain barrier permeability. Conclusion: Results suggest that paracetamol may induce an immune-inflammatory-like response in placenta and more caution should be exercised in use of paracetamol in pregnancy.


Introduction
Paracetamol (acetaminophen) is commonly taken, either by prescription or self-medication, for the relief of pain and fever. In pregnancy it is the most widely consumed drug, with estimates of expectant mothers talking this medication ranging from 56% in Australia and the Americas (Wyszynski & Shields, 2016) to 76% in Europe (Dreyer et al., 2015). The Australian Medicines Handbook (2019) states without qualification that paracetamol is safe for use in pregnancy and breast-feeding. However, epidemiological reports of behavioural effects in the offspring of mothers who took paracetamol during pregnancy are beginning to be published, suggesting a more cautious approach would be appropriate (see Bauer et al., 2018 and Discussion).
In a recent study we have found that paracetamol, when administered to a pregnant dam at doses within the clinical range used in patients, transfers across the placenta to reach the fetus at about 40% of the levels of the drug in the maternal circulation (Koehn et al., 2019b). Thus, the placenta provides a degree of protection for the developing fetus but the mechanisms involved are not yet understood, nor are the effects that paracetamol may have on placental functions. We have therefore carried out an RNA sequencing (RNA-Seq) study of E19 placentas and brains from control (un-treated) rats and from rats treated with a single (acute) or multiple (chronic) doses of paracetamol. This RNA-Seq study yielded the unexpected outcome of widespread up-regulation of inflammatory and immunerelated genes in the placenta of the dam exposed to paracetamol over a prolonged period, with a much less pronounced effect on inflammatory-related genes following a single dose; however, many other genes showed a regulatory response following a single dose of paracetamol. Inflammatory responses during pregnancy have been linked to a range of clinical complications including pre-term birth, fetal cardiac conditions and neurological deficiencies (Challis et al., 2009;Fleiss et al., 2020;Huleihel et al., 2004;Romero et al., 2007;Salafia et al., 1989). High cytokine levels in blood have been linked to increased blood-brain barrier permeability (Anthony et al., 1997;Stolp et al., 2005a) and possibly leading to a range of health complications (Brochu et al., 2011;Nelson et al., 1998;Thornton et al., 2017). Inflammation in the placenta has also been linked to increased placental permeability, as shown in studies that identified a size-dependent increase in maternal-fetal nanoparticle transfer in mice (Tian et al., 2013).
In the present study, the inflammatory response in the placenta and the fetal brain following maternal paracetamol exposure was examined to see if it was associated with alterations in placental and blood-brain barrier permeability. Placental permeability was assessed using a low-molecular weight, hydrophobic molecule sucrose to determine the transfer in both directions: from the dam's circulation to the fetal circulation and from the fetal circulation back to the dam. Transfer of a large molecule, the endogenous fetal-derived plasma protein α-fetoprotein (AFP), across the placenta into maternal circulation was also investigated. Results from both of these markers indicate that placental transfer was potentially affected by paracetamol treatment, and demonstrated increased levels of AFP detected in blood plasma of dams treated with paracetamol. The inflammatory cytokine IL-1bß was measured in fetal and maternal plasma; it showed higher levels only in fetal plasma following maternal paracetamol treatment. Permeability of the fetal blood-brain barrier to both small (sucrose) and large (plasma protein) molecules was not affected in spite of increased IL-1ß levels in fetal circulation. The results presented here highlight responses to paracetamol use during pregnancy that appear to be tissue-specific Amendments from Version 1 Statistics: Added "We also tested our data using ANOVA followed by Tukey's posthoc test. The significance levels are the same as those we obtained with a t-test." Table 1: Colour coding added, explanation in legend. Table 2: Legend: Added "(p<0.05, see Methods)." Revised as suggested by Reviewer #2 re inflammatory immune-related genes. "Symbol -indicates no significant difference in fold changes; it does not indicate that there was no fold change. This table includes only genes with inflammatory and immunerelated functions and thus does not include some of the highly expressed genes in Table 1." Table 3: Colour coding added, explanation in legend.
The effect of paracetamol exposure on placental gene expression (transcriptomic analysis): Number of upregulated genes corrected to 64. P-value added.

IL1ß concentration (ELISA):
Clarification/correction of IL1ß results: "None of the dams in any of the treatment groups had a detectable level of IL1ß in their plasma (limit <5pg/ml) nor was IL1ß detected nor was IL1ß detected in the control untreated fetuses. In contrast, IL1ß in the plasma of many of the E19 fetuses whose mothers had been treated with paracetamol was detected. The levels were generally higher in fetuses of mothers treated chronically (acute 2/4, chronic low 7/16 and chronic high 10/19)." The effect of paracetamol exposure on E19 fetal brain gene expression (transcriptomic analysis): Added "Following both treatments 26/50 genes were up-regulated and 40/50 were down-regulated."  Figure 8A. Figure 8A revised to correct error in number of controls; also corrected in legend and text: "Treatment groups were: control (1/1b; n=2), acute (2/2b; n=1),…..". Any further responses from the reviewers can be found at the end of the article REVISED and dependent on duration of treatment. The results are discussed in the context of the appropriate use of paracetamol during pregnancy.

Ethical statement
The animals used in this study were the Sprague Dawley strain of Rattus norvegicus. All animal experimentation was approved by the University of Melbourne Animal Ethics Committee (Ethics Permission AEC: 1714344.1) and conducted in compliance with Australian National Health and Medical Research Guidelines. All animals were assessed as healthy prior to commencement of experiments. Animals were monitored prior to and following every injection ensuring there was no abnormalities in weight (>10%), appearance (fur) or behaviour (vocalisation, respiration, movements). All efforts were made to ameliorate any suffering of animals. They were handled by experienced researchers in such a way as to minimise stress prior to being anaesthetised.

Animals
These were supplied by the University of Melbourne Biological Research Facility and subjected to a 12 hour light/dark cycle with ad libitum access to food (dry pellets of a fixed formulation diet for laboratory rats and mice fortified with vitamins and minerals to meet the requirements of breeding animals after the diet is autoclaved or irradiated, supplied by Speciality Feeds, Western Australia) and water. Animals were housed in groups of 2-4 (adult) per cage (25cm × 35cm × 25cm on Breeders Choice paper bedding, made from 99% recycled paper; it is biodegradable with no added chemicals). Age groups investigated (at treatment completion) were embryonic day 19 (E19) pups of both sexes and dams, which were all primigravida 350-400g body weight) and non-pregnant female adults (175-230g body weight). E19 was chosen because this is a stage of development when adequate volumes of blood and cerebrospinal fluid (CSF) can be obtained for analysis from fetal rats without pooling (Dziegielewska et al., 1981) and individual pups can be injected intraperitoneally while still inside the uterine horn and kept viable for periods of time. Animal numbers were based on previous experience of such experiments and were the minimum number required to detect a significant difference between groups at p <0.05. Animals were selected for treatment groups to ensure weights were statistically similar between direct comparisons. Where possible, equal numbers of male and female fetuses were used. Animals on gestational day E19 were allocated to experiments by animal house staff, who had no knowledge of the particular experiments to be performed. The experimenters had no role in the selection of the animals, thus avoiding selection bias. The numbers (n) of animals used for each experiment are indicated in the relevant Methods or Results section and where appropriate in legends. Two litters in the sucrose permeability studies were excluded from the study. One mother died under anaesthesia. In the other case the fetuses were observed to be in poor physiological state, which would have affected the results.

Drugs and markers
Paracetamol (acetaminophen ≥99.0%, Sigma-Aldrich) was applied either at a high dose of 15mg/kg (higher limit in the range used clinically, Australian Medicines Handbook, 2019 andKoehn et al., 2019b) or a dose in the lower clinical range of 3.75mg/kg. Paracetamol was dissolved in sterile 0.9% sodium chloride solution for injection. For passive permeability experiments [U-14 C]-labelled sucrose (Amersham International, CFB146) was injected in sterile 0.9% sodium chloride solution. Details are described in our previous study (Koehn et al., 2019b). Estimates of protein (AFP) permeability were obtained from western blot analysis of fetal and maternal plasma, as described below.
Transcriptomic analysis: RNA-Seq All experiments took place between 09.00 and 15.00h. Placentas and fetal brains from dams subjected to three treatment regimes were analysed in this study (n=4 for each tissue from each dam).
(i) an E15 pregnant dam was given an intraperitoneal (i.p.) injection twice daily with 15mg/kg of paracetamol (dissolved in sterile 0.9% sodium chloride solution) over four days. On the 5 th day (E19) the dam was given a final injection of the drug. This experiment is referred to as "chronic"; (ii) an E19 dam was given a single i.p injection of 15mg/ml paracetamol and is referred to as "acute"; and (iii) an E19 untreated dam (referred as control).
In experiments (i) and (ii), 30 minutes after the last injection of the drug the tissue samples (placentas, fetal brains) were collected (n=4 for each dam).
For RNA-Seq analysis, placental tissue was sampled as a cross section of the chorio-allantoic placental disc, following removal of the externally attached umbilical and maternal circulatory vessels. Brain samples of the cortex were dissected out as described before (Koehn et al., 2019a). Samples were collected under RNase free conditions and immediately frozen in liquid nitrogen and transferred to -80 °C for storage. RNA extraction was completed using the RNeasy Plus Mini Kits and QIAshredder (Qiagen, catalogue number 74134) for placenta and using the RNeasy Plus Micro Kits (Qiagen, catalogue number 74004) for fetal cortex, following manufacturers specifications. RNA quantity and purity were determined using a NanoDrop ND-1000 UV-VIS spectrophotometer (Thermo Scientific).
RNA samples were transported on dry ice to the Australian Genome Research Facility (AGRF) in Melbourne for Illumina, next-generation sequencing. Runs were 100bp single reads, providing raw FASTAq data. Data were processed using the Galaxy platform and their online software packages (Afgan et al., 2018). Default parameters were used unless directly specified. Alignment was conducted using HISAT2 (Galaxy version 2.1.0) using the reference genome for rat (rn6; accession number GCA_000001895.4) and the reverse strand setting. For transcript quantification and differential expression analysis, three different methods were employed. In the first, pathway transcripts were assembled with cufflinks (Galaxy version 2.2.1.2) using the reference annotation for rat RefGene (genome) obtained from UCSC Main. Relevant data were passed through Cuffmerge (Galaxy version 2.2.1.1) and analysed for differential expression between groups of interests using Cuffdiff (Galaxy version 2.2.1.5). For the second and third pathway, counts were aligned using HTSeq-counts (Galaxy version 0.9.1) using the reverse strand setting. Generate Count Matrix (Galaxy Australia version 1.0) produced a matrix form of the data, which were then fed through either DEseq2 (Galaxy version 2.11.40.6) or EdgeR (likelihood ratio; Galaxy version 3.24.1) to receive differential expression analysis between treatment groups. Statistically different expression levels between relevant treatment groups were selected if present in two of the three datasets above the statistical threshold of p <0.05 for the adjusted P value of Cuffdiff (Padj), DEseq2 (q value) or EdgeR (FDR). This method of statistical selection minimizes the known false positives and false negatives that can be obtained due to the analysis pathway selected, ensuring all results can be found between multiple pipelines (see Seyednasrollah et al., 2016;Soneson & Delorenzi, 2013). Gene synonym names were produced via bioDBnet (Mudunuri et al., 2009). Pathway analysis was conducted using DAVID Bioinformatics Resources (version 6.8), with benjamini false discovery rate correction (Huang et al., 2009a;Huang et al., 2009b).
Interleukin 1β (IL1β) enzyme-linked immunosorbent assay (ELISA) IL1β cytokine concentrations in rat plasma were determined using ELISA specific for rat IL1β (R&D systems, Quantikine kit, catalogue number RDSRLB00, monoclonal mouse anti-rat IL1β) following the manufacturer's protocol. Plasma samples were diluted 1:2 and 50µL of each sample was added to the same volume of assay diluent. Standard dilutions were assayed in duplicate. The plate was incubated at room temperature for two hours, then washed extensively. 100µl of rat IL1β conjugate was added and incubated for a further two hours. After additional thorough washes, the plate was incubated for 30minutes in 100µL of substrate solution then developed with 100µl of stop solution. Plates were read using a FlexStation 3 Multimode Microplate Reader (wavelength 450nm, using 570nm to correct for any optical imperfections in the plate) within 30 minutes of the addition of the stop solution. Cytokine concentrations were determined by comparison with the standard curve produced from each run.
Permeability across the placenta All permeability experiments were conducted on E19 dams and fetuses. Two chronic paracetamol treatment regimes were used. Time-mated E15 pregnant dams were injected i.p. twice daily with either a 15mg/kg (referred to as "high") or 3.75mg/kg (referred to as "low") dose of paracetamol (dissolved in sterile 0.9% sodium chloride solution) over four days ("chronic" experiments). On the 5 th day, at E19, these were compared to age-matched animals that were not treated (controls). Numbers (n) of individual experiments are indicated below and are included in the legends of corresponding figures in the Results section. 14 C-sucrose permeability. Animals were treated either with a "low" dose (3.75mg/kg) or "high" dose (15mg/kg) of paracetamol over four days starting at E15 following the same protocol as above. On the 5 th day the pregnant dams (E19) were anaesthetised i.p. with 25% w/v urethane, (Sigma, 1ml per 100g body weight) and placed supine on a 35°C heating plate and an endotracheal cannula inserted prior to sampling. Left femoral artery and vein were cannulated. All injections were by slow infusion to the femoral vein; the cannula was flushed with 2ml of heparinized (Hospira Inc, 5000 units per ml) saline. Maternal blood samples were taken from the femoral artery; blood volume was maintained by intraarterial injection of equivalent volumes of 1ml heparinized sodium chloride solution. Blood (right cardiac ventricle), CSF (cisterna magna) and brains (cortex) were sampled from each fetus. Sampling was concluded when the state of the placental circulation (normal condition: umbilical veins pink with oxygenated blood) was deemed insufficient, usually around 90 minutes (see Koehn et al., 2019b for details). CSF samples were examined microscopically for traces of red blood cells and discarded if contaminated (Habgood et al., 1992). Maternal blood was also collected at the end of the experiment. Blood samples were centrifuged (5000rpm, five minutes). Plasma supernatant was removed and stored at -20°C until used.
Two sets of permeability experiments were conducted: (i) Fetal to maternal placental barrier permeability: pregnant animals treated with paracetamol as described above and control, untreated dams were terminally anaesthetized and an arterial cannula inserted into maternal circulation. Once the uterine horns were exposed, individual fetuses still within their amniotic sacs were injected serially with 14 C-sucrose as described in Koehn et al. (2019b). Each fetus was taken at 30 minutes post injection. Maternal blood samples were collected at the same time as fetuses were consecutively removed for blood sampling. Maternal to fetal plasma levels ratios of 14 C-sucrose were used as a measure of fetal to maternal placental transfer and calculated as follows: One control litter (n=6); one litter from a chronically treated dam with a low dose 3.75mg/kg (n=5) and two litters from two chronically treated dams with a high dose (15mg/kg, n=5 for each litter) were used.
(ii) Maternal to fetal placental barrier permeability: pregnant animals treated with paracetamol as described above and control untreated dams were terminally anaesthetized and 14 C-sucrose was infused into the maternal circulation as detailed for paracetamol permeability studies above. Fetal samples were taken serially between 30 minutes and 90 minutes post injection. Blood samples from individual fetuses were collected together with time-matched maternal blood samples (Koehn et al., 2019b) and processed for liquid scintillation counting (see below) to obtain fetal/maternal plasma concentration ratios using the equation: One control litter (n=8) and one litter from a chronically treated dam with high dose (15mg/kg, n=6) were used.
Permeability of a fetal specific protein, AFP-western blotting. Levels of AFP in both the maternal blood samples and in fetal samples obtained from experiments of paracetamol treated dams as described above, were estimated using western blotting and antibodies to human AFP (DAKO).
All plasma samples were diluted 10-fold in isotonic saline (0.9%) prior to sample preparation. Samples were run using a total of 9µL of dam and 2µL of diluted fetal sera, denatured in 4x sample buffer (62.2 mM Tris, 5% (v/v) glycerol, 2% (w/v) SDS, and 0.0025% (w/v) bromophenol blue), heated to 95°C for five minutes and centrifuged briefly to remove potential particular matter. 12µL of each sample was loaded onto a 4-12% NuPAGE Novex Bis-Tris Midi gel (Life Technologies) and proteins were resolved at 200V for 40 minutes immersed in MES SDS running buffer (Life Technologies). Gel-resolved proteins were transferred onto PVDF membranes using iBlot gel transfer stacks (iBlot 2; Life Technologies) as per manufacturer's instructions. Membranes were incubated for one hour at room temperature in PBS-T blocking buffer (PBS supplemented with 0.05% (v/v) Tween-20 [Chemsupply]) and 5% (w/v) skim milk powder. Membrane was incubated with AFP primary antibody (AFP, rabbit polyclonal, 1:1000, DAKO, catalogue number A0008, RRID AB_2650473) diluted in the blocking buffer and incubated overnight at 4°C. After three PBS-T washes, the membrane was incubated for two hours at room temperature in horseradish peroxidase-conjugated goat anti-rabbit IgG (Cell Signaling; 1:5000, catalogue number 7074) secondary antibody. Immunoreactive protein bands were visualised by adding 1mL of Enhanced Chemiluminescence mixture (ECL Advance, GE Healthcare) onto membranes and detecting luminescence using a FujiFilm LAS-3000 imager at three and 75 second exposures. Densitometric quantitation of immunoreactivity was performed using ImageJ 2-bit, v1.46 run on OSx 10.14 Mojave on 8-bit TIFF file images. All samples that were directly compared were run on the same gel. Serum from an age-matched non-pregnant female was used as a negative control, while an E19 pregnant dam that was not injected with paracetamol was used as a positive control. Both samples were included on every gel.
Permeability of the fetal blood-brain barrier Blood-brain barrier permeability in the fetus was estimated using (i) radioactive sucrose as an example of a small molecular marker permeability and (ii) plasma protein (immunohistochemistry), as an example of a large molecular marker permeability (Habgood et al., 1993;Johansson et al., 2008;Stolp et al., 2005b). Fetal blood, CSF and brain samples were obtained from the same placental permeability experiments described above. 14 C-sucrose permeability. For estimation of transfer from mother to fetus, pregnant animals treated with paracetamol (as above) were anaesthetized i.p. with urethane. Starting at 30 minutes after the last maternal injection, embryos were individually extracted. For estimation of transfer from fetal blood to fetal brain and CSF, the fetuses were exposed and injected i.p. with 14 C-sucrose.
In both types of experiment fetal blood and CSF were sampled as described previously (Koehn et al., 2019b). Fetal brain samples were taken by opening the cerebral hemispheres to expose the lateral ventricles and a sample of the parietal cortex was removed, taking care to avoid the choroid plexuses. Brain or CSF to plasma ratios of 14 C-sucrose radioactive counts were used as an estimate of the transfer of sucrose across the blood brain barriers. These were calculated using the equation: Treatment groups investigated were control, no paracetamol (n=13), chronic low dose (3.75mg/kg, n=11) and chronic high dose (15mg/kg, n=11) in fetuses that were injected directly. In experiments in which the 14 C-sucrose was injected into the treated mothers, numbers of pups used were control (n=8), acute (n=10) and chronic high dose (n=6).
Immunohistochemistry. Individual fetal brains were fixed in Bouin's fixative for 24-48h then dehydrated through graded alcohols, cleared in chloroform and embedded into paraffin wax blocks. These blocks were cut serially into coronal 5µm sections (Leica microtome). Selected sections were heated for 30 minutes (60°C) then washed twice with Histolene (Fronine) for 10 minutes, then five minutes. The sections were rehydrated through graded alcohols for five minutes each (100%, 100%, 95%, 70%) and washed in phosphate buffered saline with 0.2% Tween20 for five minutes. Peroxidase and protein blockers (DAKO) were added to sections and incubated at room temperature for two hours each to block non-specific binding. The primary antibody, plasma protein (anti-rat whole serum, SIGMA, catalogue number R5129, rabbit polyclonal) diluted 1:3000 in a blocker (0.5% fish gelatine and PBS + Tween20), was applied to the slides and incubated overnight at 4°C. After three washes of PBS + Tween20 for five minutes each, the secondary (swine anti-rabbit, DAKO, catalogue number Z0196, polyclonal) and tertiary antibodies (rabbit PAP, SIGMA, catalogue number P1291) both diluted 1:200 were each added and incubated for two hours at room temperature with washes between incubations. Sections were developed with DAB (Diaminobenzidine) using DAKO DAB+ kit (catalogue number K3468) according to manufacturer's directions and washed in running water for five minutes. Sections were dehydrated through a series of graded alcohols (70%, 95% for five minutes, then 100% for 10 minutes), then 3x five minutes in histolene washes. Slides were then mounted using DPX mounting medium (Fronine). Stained sections were examined under a compound microscope (Olympus, BX50) fitted with a digital camera (Olympus DP70). One control slide was included with every round of immunostaining and had the primary antibody omitted but was otherwise treated in the same way. These were always blank. A total of 11 brains with at least two brains per treatment group were prepared and serially sectioned and mounted on glass slides. Each slide contained 6-8 sections, every 10th slide was stained with haematoxylin and eosin for general morphology. One or two adjacent slides per brain were immunostained for plasma protein from comparable brain regions.
Liquid scintillation counting Plasma (10µL), CSF and every injectate (1µL of 1:10 dilution) were weighed and transferred into scintillation vials. In all experiments the radioactivity in the injectate was measured to confirm the uniformity of the injected material. Soluene350 (0.5ml, PerkinElmer) was added to the brain samples and incubated overnight at 36°C. Prior to measurement, two drops of glacial acetic acid (Sigma) were added to brain vials to neutralize the strongly alkaline Soluene350. All samples were mixed with 5ml of scintillation fluid (Emulsifier-safe, PerkinElmer) and measured on the liquid scintillation counter (Tri-Carb 4910 TR, PerkinElmer). Counting was conducted in disintegrations per minute (DPM) for five minutes each with luminescence correction on. Vials containing control, non-radioactive tissues processed identically were also counted simultaneously to establish background counts (which were subtracted from all radioactive samples). Counts were normalized to the sample weight and expressed as DPM per µL or µg of sample. Results are described as concentration ratios, defined as a % of the counts (per µL or µg) in the compartment of interest (brain, CSF, maternal or fetal plasma) divided by the counts (per µL) in the plasmacompartment of comparison (see also Koehn et al., 2019b).

Statistics
RNA-Seq data analysis is detailed above, with significance set at p <0.05. For all other experimentation, statistical differences between treatment groups were determined by unpaired Student t-tests using Prism 6.2 (Graphpad Software Inc) with significance set at p <0.05. We also tested our data using ANOVA followed by Tukey's posthoc test; this approach yielded the same outcomes.

Results
E19 placentas and brains from three treatment groups were compared for transcriptomic analysis using RNAseq datasets: (i) untreated controls (n=4), (ii) acutely paracetamol treated (n=4) and (iii) chronically paracetamol treated (n=4) dams (see Methods), providing a three-way comparison for each tissue ( Figure 1 and Table 1-Table 5).  . Colours indicate genes that were upregulated (green) in both acute and chronic treated animals and downregulated (red) in both acute and chronic treated animals. Note that only 10/50 genes were upregulated following both treatments but 34/50 were downregulated following both treatments. Table 2. Changes in gene regulation in E19 placenta following maternal treatment with paracetamol.

ID Gene
Ch / Co

Ac / Co
Ch / Ac

-
Up-regulated and down-regulated inflammatory and immune-related gene changes in E19 placenta following no treatment (co, controls), acute (ac,) or chronic (ch) maternal paracetamol treatment; see Methods for details of dosage schedule. Data from RNA-Seq analysis. Numbers are fold changes for comparisons indicated (Ch/Co, Ac/Co, Ch/Ac). The chronic/acute comparison indicates significant differences in regulation between the two dosage regimes (P<0.05, see Methods). In all cases expression was greater with chronic treatment. This table includes only genes with inflammatory and immune-related functions and thus includes some of the highly expressed genes in Table 1.
The effect of paracetamol exposure on placental gene expression (transcriptomic analysis) As illustrated in Figure 1, following maternal exposure to paracetamol (either acute or chronic), there was a large number of genes that significantly altered their expression in the E19 placentas in two-way comparisons to control tissue, with much fewer that changed between the two treatment groups (chronic/ acute). Most genes were uniquely regulated, either up or down, depending on treatment duration, with relatively few that were common to both treatment regimes (64 up-regulated and 57 down-regulated). In contrast, in a three-way comparison, only one gene, Nfkbia (NF-kappa-B inhibitor alpha), was shared in all three comparisons (Figure 1). NFKB is a transcription regulator that is activated by various intra-and extra-cellular stimuli such as cytokines.
The expression of 121 transcripts (the sum of up-regulated and down-regulated genes in the chronic/acute comparison) was significantly different between acute and chronic treatment groups, suggesting an effect of treatment duration. Of these genes, 34 were significantly up-regulated in chronically treated animals when compared to either the acute treatment group or the control group and eight were down-regulated ( Figure 1).
Comparing datasets of placentas from chronically treated dams with untreated control dams, the expression of 737 genes was significantly different (either up or down p<0.05, see Methods) (Figure 2). The top 50 up-regulated and down-regulated genes in E19 placentas are displayed for both acute and chronic treatment groups compared to controls in Table 1. In the E19 placentas, many of the top genes up-regulated following chronic treatment were related to immune-response and inflammation (Table 1, Table 2 and Table 5). It is difficult to determine the extent to which a statistically difference in gene expression is also functionally significant. It is perhaps worth noting that fewer genes were up-regulated two-fold or more with either acute or chronic paracetamol treatment (25 and 34 genes, respectively) compared to the number that were down-regulated two-fold or more (58 and 61, respectively). In addition, the degree of down-regulation was appreciably greater for many of these genes compared with those that were up-regulated. This was particularly evident for the chronically treated group Fold change of transcript numbers following chronic, acute or control maternal treatment with paracetamol, n=4 in each group. For details of dosage schedules see Methods. Data from RNA-Seq analysis. FC = fold change compared to control. Colours indicate genes that were upregulated (green) in both acute and chronic treated animals and downregulated (red) in both acute and chronic treated animals. Note that 26/50 genes were upregulated following both treatments and 40/50 were downregulated following both treatments. Table 4. Changes in gene regulation in E19 fetal brain following maternal treatment with paracetamol.

ID Gene
Ch / Co Ac / Co Ch / Ac ID Gene Ch / Co Ac / Co Ch / Ac

ID Gene
Ch / Co Ac / Co Ch / Ac ID Gene Ch / Co Ac / Co Ch / Ac Up-regulated and down-regulated inflammatory and immune-related gene changes in the E19 fetal brain following no treatment (Co, controls), acute (Ac,) or chronic (Ch) maternal paracetamol treatment; see Methods for details of dosage schedule. Data from RNA-Seq analysis. Numbers are fold changes for comparisons indicated (Ch/Co, Ac/Co; Ch/Ac). Compared to the placenta, in the fetal brain many fewer inflammatory and immunerelated genes showed regulatory changes and there were no significant differences between acute and chronic treatments. -indicates no significant difference in fold changes, not that there was no fold change.
compared to the control group, with five genes down-regulated greater than 500-fold (Afp, apoc2, rbp4, apob and fgb, Table 1). In addition, 10/50 genes were up-regulated following both treatments but 34/50 were down-regulated following both treatments. Thus overall the down-regulatory effects of paracetamol were much more pronounced than the up-regulatory effects.
Genes that showed a regulatory response in placentas of animals following both acute and chronic treatment with paracetamol are listed in Table 5. Seven of these down-regulated genes showed a fold change of more than two, which was greater in the chronically treated placentas. Other changes were so small that they are unlikely to be of much functional significance.

The inflammatory response
In the placenta of chronically treated rats there was a notable up-regulation of immune response related genes compared to the acutely treated group (Table 2). Figure 3 illustrates an analysis from biological Gene Ontology (GO) categories of immune response genes (A) subdivided into the innate (B) and adaptive (C) immune systems in the chronically treated animals. In the placenta these included GO biological processes such as neutrophil chemotaxis (p=4.7E-05) and innate immune response (p=0.045). Figure 3 illustrates that the number of significantly up-regulated genes was substantially more than the number of down-regulated genes and that most of these were in the innate immune system category, with a small number in the adaptive immune system. A list of inflammatory and immunerelated genes that were up-regulated in the placenta following chronic treatment is shown in Table 2. Overall, some 36 genes showed a statistically significant up-regulation. These included 15 genes that were up-regulated two-fold or more. As can be seen from Table 2, the third most up-regulated gene in the placenta following chronic paracetamol exposure was Il1ß. Figure 4 illustrates the number of Il1ß gene transcripts in the three treatment groups in the fetal brain and placenta. There was a prominent increase in Il1ß transcripts in the placentas from the dams treated with chronic paracetamol and no difference between the datasets of placentas from the control and Table 5. Inflammatory and immune-related gene regulation in both acute and chronic treatment with paracetamol. Only genes that showed a response in placentas from E19 animals (left panels) and fetal brains (right panels) following both acute and chronic maternal treatment with paracetamol are shown; see Methods for details of dosage schedule. Data from RNA-Seq analysis. Numbers are fold changes for comparisons indicated (Ch/Co, Ac/Co; Ch/Ac). There were no significant differences for these genes between acute and chronic treatments, although there were small fold changes (data not shown).  acutely treated mothers, both showing very low numbers.

ID
IL1ß is a prototypical marker for inflammation and immune response, with up-regulation in the chronically treated placenta of 13.3 fold change; it could thus be a potential indicator of fetal harm. The response in the placenta following a single acute dose of paracetamol was much more muted (Table 2).
Amongst the down-regulated genes in the placenta (Table 1) were several transcripts for plasma proteins (AFP, transthyretin and transferrin, see Discussion) that have been shown to downregulate under inflammatory conditions (negative acute phase response, Heinrich et al., 1990;Hu et al., 2019;Mackiewicz et al., 1990). Two of these were markedly down-regulated in the acute experiments and further down-regulated in the chronic experiments (Table 1). This suggest that the response of these plasma protein genes was rapid in onset and continuing over several days in the presence of chronic treatment. In contrast, the up-regulatory response of cytokine genes only became prominent in the placentas of animals chronically exposed to paracetamol (Table 1; Figure 4).

IL1ß concentration (ELISA).
In order to see if the increase in transcript numbers for IL1ß in placentas from dams treated chronically with paracetamol ( Figure 4) translated into an increase in its protein concentration, the levels of this cytokine in plasma of both the dams and pups were measured using a commercially available ELISA kit (see Methods). Results are illustrated in Figure 5. None of the dams in any of the treatment groups had a detectable level of IL1ß in their plasma (limit <5pg/ml) nor was IL1ß detected in the control untreated fetuses. In contrast, IL1ß in the plasma of many of the E19 fetuses whose mothers had been treated with paracetamol was detected. The levels were generally higher in fetuses of mothers treated chronically (acute 2/4, chronic low 7/16 and chronic high 10/19).
The effect of paracetamol exposure on E19 fetal brain gene expression (transcriptomic analysis) Transcriptomic analysis of the E19 fetal brain was carried out in material collected from the same animals as was prepared for placental analysis, thus allowing a direct comparison between the response of the two tissues to paracetamol treatment of the mother.
As illustrated in Figure 1, following maternal exposure to paracetamol, there was a large number of genes that significantly altered their expression in the fetal brain.
As shown in Figure 2, comparing the dataset for fetal brains from chronically treated dams with untreated control dams, there was a total 1128 genes with significantly different transcript numbers in the E19 brain. The top 50 up-regulated and down-regulated genes in the E19 brain are shown for both acute and chronic treatment groups compared to controls in Table 3. Following both treatments 26/50 genes were up-regulated and 40/50 were down-regulated. Additionally, the level of  down-regulation was greater for most transcripts than up-regulation following both acute and chronic paracetamol treatment, for example Col1a1 (collagen type 1 alpha 1 chain) and Col3a1 (collagen type 3 alpha 1 chain), see Table 3. There will be a further analysis of the brain data in a later publication (Koehn et al., unpublished reports) that will deal with expression of ABC efflux transporters and related enzymes as these may play a role in the extent to which paracetamol enters the brain at different stages of development (Koehn et al., 2019b).
Comparison of the inflammatory response in E19 placenta and brain following maternal paracetamol treatment In addition to effects of the length of exposure to the drug on gene expression in individual tissues, the regulation in brain and placenta was very different following the same treatment.
Only 98 genes were significantly regulated in both tissues, equating to 5.5% of the transcripts that changed their expression (Figure 2).
In the E19 placenta many of the top genes up-regulated following chronic treatment were related to immune-response and inflammation, including Il1ß, which was 3 rd highest (Table 1). In contrast, in the brain, very few transcripts for Il1ß (Figure 4) or other cytokines (Table 4) could be detected and there was no difference in transcripts for Il1ß between the treatment groups ( Figure 4). Table 4 lists the inflammatory and immunerelated genes that were up-or down-regulated significantly in the E19 brain. The changes were very small compared to the placenta in both the innate immune and the adaptive immune category (Figure 3). Table 5 shows immune/inflammatory related genes that showed a regulatory change in both the placenta and the brains from both acutely and chronically treated fetuses.
No changes in plasma protein transcript numbers were detected in the fetal brain (see Discussion). This, together with lack of up-regulation of the inflammatory cytokine Il1ß, as seen in the placentas, indicates that an inflammatory response was elicited by paracetamol in the placenta but little or none in the fetal brain. We do not have information if other organs not investigated in this study, such as the liver, could also have been affected.

Placental permeability
In order to investigate if a prolonged exposure to paracetamol and resulting inflammatory response could affect the permeability of the placenta, two sets of permeability experiments were conducted using a small molecular size marker, 14 C-sucrose (see Methods). These were designed to examine the transfer from the mother to the fetus but also from the fetus back to the mother. Results are illustrated in Figure 6 and Figure 7.

Fetal to maternal transfer of 14 C-sucrose.
To investigate the placental transfer of sucrose from fetus back to the dam following maternal paracetamol exposure, sucrose was injected directly into the pups still within their amniotic sacs (see Methods). Two litters were injected in mothers that had been treated with chronic high doses of paracetamol and one litter from a mother treated with chronic low dose paracetamol. These were compared with one litter from an untreated control mother. Plasma samples from both the fetuses and dam were collected and ratio of 14 C-sucrose estimated (see Methods). The results are shown in Figure 6. All three of the litters from mothers treated with chronic paracetamol (either high or low dose) showed slightly higher permeability from the fetus back to the mother than in the control dam. However, the ratios are extremely low, making accurate comparison difficult.

Maternal to fetal transfer of 14 C-sucrose.
In order to investigate if the rate of transfer of a small molecular marker from dam to fetus across the placental barrier was affected following paracetamol exposure, dams either untreated (control) or treated with chronic high (15mg/kg) doses of paracetamol were given a final intravenous (i.v.) injection of 14 C-sucrose 30 minutes before removing their fetuses (Figure 7). Blood samples from dams were time matched to the removal and blood collection from each fetus (see Koehn et al., 2019b). The transfer from the mother to the fetus in the paracetamol treated dams was slightly less than that in the control animal. The much higher ratios obtained in the maternal to fetal transfer experiments (Figure 7) Figure 7. Maternal to fetal (E19) transfer of 14 C-sucrose. Mothers were treated with paracetamol. 14 C-sucrose was injected (i.p) into the mothers. Blood samples from individual fetuses were collected at the same time as maternal samples. Treatment groups were: untreated (control, n=8) and paracetamol injected (chronic dose 15mg/kg, n=6) dams. n refers to number of fetuses. Each data point is a single fetus. Transfer calculated as fetal/maternal plasma ratio (%).

Figure 6. Fetal (E19) to maternal transfer of 14 C-sucrose.
Following maternal paracetamol treatment fetuses, still within their amniotic sacs, were directly injected (i.p.) with 14 C-sucrose and plasma from both fetus and dam were collected to calculate maternal/average fetal plasma ratio (%) over time. Treatment groups: control (n=6), chronic low dose (3.75mg/kg, n=5) and chronic high dose (15mg/kg, n=5); n refers to number of fetuses. Each data point is a single fetus.
compared to the fetal to maternal transfer ( Figure 6) are due to the differences in volume of distribution, hence dilution of the marker, when sucrose is injected into the mother or into the fetuses.

Detection of AFP in fetal and maternal plasma.
In order to investigate if exposure to paracetamol can also influence the transfer of a protein from the fetal circulation into the maternal blood across the placenta, western blot analysis was made of fetal and maternal plasma samples using cross-reacting antibodies specific for AFP (see Methods). Figure 8A shows the blot that contained both the fetal and maternal samples together with one negative control (non-pregnant female rat). Densitometry measurements are illustrated in Figure 8B together with maternal/fetal ratios. There was no detectable band in the non-pregnant control sample and all maternal samples showed a much lower level of the protein than fetal samples. The levels of the protein in fetal samples did not appear to change between the control and any of the treatment groups ( Figure 8B), but in maternal samples, AFP levels were higher in all chronically treated dams compared to un-treated controls. This was reflected in the ratios of AFP in maternal to fetal plasma ( Figure 8B, right panel) in which all of the chronically treated animals had ratios that were above those in untreated controls and in one acutely treated animal. Prolonged exposure to the drug increased AFP transfer from fetus to dam by about three times compared to the control animals.
Permeability of the fetal blood brain barrier Two different molecular size markers ( 14 C-sucrose and plasma proteins) were used to assess any changes in blood-brain barrier permeability following chronic paracetamol treatment of the dam. The samples were obtained from the same experiments as the placental permeability studies.

Transfer of 14 C-sucrose into the brain and CSF following different paracetamol treatment regimes.
To investigate the transfer of 14 C-sucrose into the fetal brain after maternal paracetamol exposure, fetal blood, brain and CSF samples from dams untreated (control) or treated acutely or chronically with either low (3.75mg/kg) or high (15mg/kg) doses of paracetamol were measured. As shown in Figure 9, there was no significant difference in the transfer into the brain and CSF between any treatment groups ( Figure 9).
Entry into the brain and CSF when the 14 C-sucrose was injected directly into the fetus was also investigated and results are illustrated in Figure 9. Here too there were no significant

Figure 8. Estimations of α-fetoprotein (AFP) concentrations in fetuses (E19) and dams. A) Western blots of AFP in plasma from
dams and fetuses in different treatment groups. Numbers for dam blots are samples from individual animals; numbers in fetal plasma blots indicate individual fetuses from corresponding dams. Treatment groups were: control (1/1b; n=2), acute (2/2b; n=1), chronic low dose (3/3b; n=2), chronic high dose (4/4b; n=3) and non-pregnant control (5; n=1). B) Estimations of AFP in dam and fetal plasma (densitometry units from blots in (A) and fetal to maternal transfer of AFP expressed as dam/plasma AFP ratio (%). Note: each point represents an individual animal. Note that all chronic treated dams had higher plasma levels of AFP than the un-treated control pregnant dams; AU are ordinate arbitrary densitometry units. Figure 9. Transfer of 14 C-sucrose into the E19 brain (A) and cerebrospinal fluid (CSF; B) following paracetamol treatment. Fetuses were exposed to 14 C-sucrose either directly (fetal i.p. injection) or indirectly (maternal i.v. injection). Treatment groups investigated were control, no paracetamol (n=13), chronic low dose (3.75mg/kg, n=11) and chronic high dose (15mg/kg, n=11) in fetuses that were injected directly. In experiments in which the 14 C-sucrose was injected into the mothers; n numbers were control (n=8), acute (n=10) and chronic high dose (n=6) in the mothers. Means ± SD.
differences in the entry of sucrose into brain and CSF between the three treatments.
However, the fetuses that were directly exposed to sucrose (i.p injection) showed a lower level of transfer into the brain and CSF compared to those that were exposed indirectly (i.v. injection to dam), around 10% compared to 40%, respectively. This reflects differences in distribution volume following the different routes of injection as well as the time involved in samples collection.

Blood brain barrier integrity for endogenous plasma protein.
Transfer of large molecule plasma proteins into the fetal brain following maternal paracetamol exposure was studied using immunohistochemistry and antibodies to rat serum proteins (see Methods). Brains were matched with plasma samples containing detectable IL1ß levels as estimated by ELISA ( Figure 5). The distribution of the proteins in E19 brains from control, acute and chronic high dose (15mg/kg) paracetamol treated dams is illustrated in micrographs in Figure 10. There was no evidence of a "leak" of protein in any of the vessels in the fetal brains examined. In all sections stained from all brains investigated, immunostaining was exclusively localised in the blood vessels, choroid plexus stroma and precipitated CSF and there was no visible difference in the brain morphology between treatment groups.
Thus, the results clearly show that the blood-brain barrier, at least to plasma proteins and to sucrose, was not affected by paracetamol exposure of the dam, the inflammatory response in the placenta nor the increased levels on IL1ß in fetal blood.

Discussion
In order for a drug taken by a pregnant mother to reach the fetal brain it has to cross both the placental and the blood-brain barriers. Any changes to normal functioning of these interfaces could have detrimental effects on fetal health and pregnancy outcomes. We have therefore analysed the transcriptomic changes in rat E19 placentas and brains following paracetamol treatment of the dams. Paracetamol is one of the most commonly used medications in pregnancy (Dreyer et al., 2015;Wyszynski & Shields, 2016). Pregnant rats were treated with paracetamol acutely and chronically and compared to controls (no treatment). The doses used were within the clinically recommended range (0.5g to 4g in 24 hours in adults). In the case of the chronic treatment, this corresponded to a relatively prolonged period of pregnancy in the rat (E15-19, about 25% of gestation). This was followed by investigating placental transfer of small and large molecules from the dam to the fetus and from the fetus back to the maternal circulation, to see whether paracetamol exposure altered barrier function. Finally, the permeability of the blood-brain barrier was analysed in the fetuses of paracetamol treated and untreated dams.
From the results it was apparent that some form of acute phase response was elicited as transcripts for several plasma proteins were down-regulated in placentas of both acute and chronic treated animals (Table 1). These proteins were AFP, transthyretin and transferrin (Vranckx et al., 1989), fibrinogen beta chain (Birch & Schreiber, 1986) and apolipoproteins Apoa1-4, several of which are known to respond to inflammation as negative acute phase proteins (Tu et al., 1987). Since a marked response was already apparent after a single dose of paracetamol, it seems that this was a rapid response to paracetamol, which was sustained and increased when the treatment was chronic. A summary of transcript numbers for AFP, transferrin and transthyretin, together with numbers for Il1ß for comparison, is presented in Table 6 for both the brain and the placenta. These clearly show that some form of acute phase response was taking place in the placenta following paracetamol treatment; however, other typical acute phase response-related cytokines were not up-regulated (such as TNFα or IL6). Transcript numbers in the brain did not change, demonstrating that the acute phase response was tissue specific and restricted to the placenta.
Several immune and inflammatory-related genes were up-regulated in placentas of animals treated with the chronic doseregime, but much less so in the placentas of acutely treated animals (Table 1, Table 2 and Table 5). The key inflammatory cytokine, IL1ß, was shown to be present in the blood of a high proportion of fetuses of mothers exposed to both acute and chronic treatment with paracetamol. The levels were variable in different fetuses but generally higher in the chronically treated animals. No IL1ß could be detected in either the maternal blood of paracetamol treated animals or in fetuses of control untreated animals. This confirms that paracetamol was indeed eliciting an inflammatory response but only on the fetal side of the placental circulation. Thiele et al. (2015) reported that pregnant mice treated with either 50 or 250mg/kg paracetamol showed some immune responses in the uterus and some morphological changes in the placenta. However, they did not investigate possible immune responses in the placenta and the doses of paracetamol were much larger than the ones we used and were well above the clinical range.
In order to determine if prolonged paracetamol exposure of the dam could affect some aspect of placental function, we have estimated placental permeability to a small molecular marker, sucrose and to large plasma protein AFP in both directions i.e. from the dam to the fetuses and from the fetuses back to the dam. The results showed that there was a small and variable increase in permeability to 14 C-sucrose and of AFP permeability in the direction from fetus to mother ( Figure 6 and Figure 8). There may also have been a small decrease in sucrose permeability from mother to fetus (Figure 7) but due to small numbers this is inconclusive.
Placental inflammation induced by lipopolysaccharide (LPS) injection in pregnant rats has been reported to induce maternal serum and placental cytokines and increased maternal serum AFP (Hu et al., 2019). In those experiments LPS did not increase the expression of AFP in fetal liver, maternal liver or placenta, but did reduce the fetal serum AFP levels, a pattern implying a possibility of increased transfer of AFP from the fetus to the mother, thus depleting it from fetal circulation. We did not find any difference in fetal AFP levels but this discrepancy could be due to either the duration and severity of the response or sensitivity of the methods used.
Permeability of the fetal blood-brain barrier to both sucrose and plasma protein was also investigated. In contrast to the placenta, there was no evidence of a change in brain barrier permeability to either marker in fetuses of dams treated with paracetamol. This is relevant to earlier studies in which inflammation induced by LPS was shown to result in a breakdown of the blood-brain barrier that was age-dependant (Stolp et al., 2005a;Stolp et al., 2005b). However, it is likely that E19 is at a developmental stage when the response to LPS is not yet developed, as shown in a similar study in a marsupial species, Monodelphis domestica (Stolp et al., 2005a).

Limitations of the study
The study has been carried out in pregnant rats at a single gestational age (E19). This stage of brain development in rats at E19 is approximately equivalent to 22-24 weeks gestation in humans (Clancy et al., 2001), corresponding to the earliest age of viability (Fischer et al., 2009;Stoll et al., 2010). The rat and human placentas are both classed as hemochorial (Blood et al., 2007;Dawe et al., 2007) but there are differences in morphology, in particular that the rat placenta has more morphological layers between the fetal and maternal circulations. However, that might mean that the relatively small changes in placental permeability from fetus to mother shown here might be more prominent in the human. The responses of these two species to an inflammatory event are similar with respect to the three plasma proteins AFP, transferrin and transthyretin (prealbumin); as in this study, these proteins have been reported to be acute phase negative proteins under inflammatory conditions (Heinrich et al., 1990;Hu et al., 2019;Mackiewicz et al., 1990). This supports the suggestion that these findings should be taken account of when advising pregnant women about the use of paracetamol. Given the unexpected findings of up-regulation of inflammatory cytokines and down-regulation of some acute phase plasma proteins, we are in the process of carrying out RNA-Seq replication studies and extending the range of cytokines estimated in fetal and maternal blood. Unfortunately, these experiments have been delayed by the COVID-19 emergency, which has In placental samples transcript counts per million for the three negative acute phase proteins were smaller in chronically treated animals compared to controls and all IL1β numbers were greater than in controls.
There was some variation in values between individual placentas which was more obvious in samples from acutely treated animal indicating that the response was potentially time-dependent. In all brain samples the transcript numbers were very low, with no evidence of an inflammatory response. This indicates that the response in the placenta was tissue specific.
closed our laboratories for an indefinite period. In view of the potential significance of our findings for the use of paracetamol in pregnancy, particularly the high frequency of its use, we feel it is justified to present these findings for peer review, in their present form.
Clinical relevance Paracetamol (acetaminophen) is generally considered "safe" to use in pregnancy and lactation (Australian Medicines Handbook, 2019;Briggs et al., 2017) although it is one of the most commonly overdosed drugs, including in pregnancy (Rayburn et al., 1984). However, some authors urge caution in its use because of evidence of adverse effects (Brune et al., 2015). It has been reported that as many as nearly 80% of pregnant women in some populations ingest paracetamol (Dreyer et al., 2015).
The findings of the present study, although based solely on experiments in rats, should be taken account of when advising pregnant patients on the use of paracetamol in pregnancy. The clinical situation is not straightforward because in addition to taking paracetamol to relieve pain, it may also be taken to reduce an increase in body temperature accompanying an infection (often respiratory), but there is evidence of an association between infection/fever and adverse outcomes for pregnancies; this seems to be a particular problem when infection/fever occurs at the beginning of the 3 rd trimester (Hagberg et al., 2015). Thus, continued but limited use of paracetamol to control severe pain and to reduce body temperature at critical stages of pregnancy would seems to be appropriate but not the widespread use for lesser indications that is implied by the reports that most pregnant women take paracetamol.
Increased transfer of sucrose and AFP from fetal circulation into maternal circulation, as demonstrated in the present study, suggests that other molecules/metabolites could potentially also reach the maternal circulation. There are several clinical implications, including that increased AFP levels detected in pregnant women are used to detect potential neural tube closure defects, although this test is done earlier in pregnancy and we have as yet no evidence of paracetamol affecting placental permeability this early in pregnancy.
Further investigation is required to see if there are similar effects in the placentas of patients who have taken paracetamol. If the effect is indeed confined to the fetal side of the placenta it will be clinically difficult to determine such an effect in pregnant patients, particularly if it turns out to be variable as in our rat experiments, although transfer of AFP from fetal to maternal circulation might be a useful indicator.

Open Peer Review
paracetamol during pregnancy. It examines gene expression changes in placenta and foetal brain and also presents out a number of functional studies to establish whether placenta permeability or brain barriers in the fetus are affected by the treatment. The study reports significant changes in the expression of a number of genes, including genes associated with the immune response, and validates changes in one of these genes, Il1ß, at the protein level in the placenta. While in some respects this study is still preliminary, the information presented here is valuable for underpinning future studies.
The authors clearly explain their choice of the end point (E19) selected, but not the reason for starting the chronic treatment at E15, rather than earlier, when important developmental events occur and teratogenic effects might be more likely and significant.
Specific Comments: P6: It is not clear why only the t-test was used when comparing multiple groups, as ANOVA followed by a post-hoc test should have been used.
On p7 the authors say "...65 up-regulated and 57 down-regulated....", but Fig. 1 indicates 64 upregulated genes, consistent wih the total of 121 up and down-regulated transcripts indicated in the right column.
The authors indicate that expression of 737 genes is significantly affected by chronic treatment, but do not show the level of significance. Does this mean that p is <0.05 (but never <0.01 or smaller) for all transcripts? Table 1 and 3. It would be helpful to colour code genes that change in both acute and chronic treatment groups and use thicker vertical lines between groups for ease of visualization. Table 2 includes genes that are not in the top 50 shown in Table 1, and this should be clearly stated (at a first glance the Table seemed a bit redundant). As for Table 1, the level of significance should be indicated. The Table could be made it easier to read if the "up-regulated (acute/control)" genes were shown below the "up-regulated (chronic/control)", rather than in adjacent columns, or were clearly separated using a thicker vertical line. In addition, it is confusing to have a column "chronic/control" under the "up-regulated (acute/control)" list. This seems to have been done to accommodate S100a rather than inserting it under each comparison. Please check carefully that the difference indicated in different Tables are the same (e.g. S100a8 has a FC 2.25 in Tab1 and FC 2.26 in Table 2). "Il1b" should be changed to "Il1ß". A pie chart of the inflammatory genes to complement Table 2 and Fig. 3 would be useful.
P17, left column, top and Fig. 4. There is clearly variability, but to give numbers of fetuses where Il1ß levels could be detected over total numbers assayed for all groups would be more accurate and informative (e.g. acute 2/4, chronic low 7/16 and chronic high 10/19) than including these number only for the chronic high group, which appears to be wrongly given as 19/39, while the number of fetuses indicated for this group in Fig. 5 legend is 19.
Figs 6 and 7 do not include error bars and no statistical analysis of these data seems to have been performed. It should be clearly indicated whether there was no statistical difference among groups at any time point studied.
P20, left column, top. The statement: "AFP levels were higher in all treated dams compared to an un-treated control." should be revised, as Fig. 8 shows an AFP increase only in chronically treated dams. It is a pity that the number of dams is too small to assess the significance of this observation and that no housekeeping protein was used to normalize AFP expression. If B is a densitometry of the gel in A, where according to the western blot labelling and the legend there is only 1 control for both dam and fetus, why are there 2 samples indicated in the controls in the charts? Given the low sample numbers and variability, particularly in fetal AFP levels, expressing the data as ratio is not appropriate.

Is the work clearly and accurately presented and does it cite the current literature? Partly
Is the study design appropriate and is the work technically sound? Yes

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly permeability or brain barriers in the fetus are affected by the treatment. The study reports significant changes in the expression of a number of genes, including genes associated with the immune response, and validates changes in one of these genes, Il1ß, at the protein level in the placenta. While in some respects this study is still preliminary, the information presented here is valuable for underpinning future studies.
The authors clearly explain their choice of the end point (E19) selected, but not the reason for starting the chronic treatment at E15, rather than earlier, when important developmental events occur and teratogenic effects might be more likely and significant. On p7 the authors say "...65 up-regulated and 57 down-regulated....", but Fig. 1 indicates 64 up-regulated genes, consistent with the total of 121 up and down-regulated transcripts indicated in the right column.
We thank the Reviewer for drawing our attention to this error which has been corrected.
The authors indicate that expression of 737 genes is significantly affected by chronic treatment, but do not show the level of significance. Does this mean that p is <0.05 (but never <0.01 or smaller) for all transcripts?
As indicated in the Methods section on "Statistical Analysis" we used P<0.05 for two of the three analyses used. We focussed on genes with large fold changes as these are more likely to be of functional significance than would be indicated by a higher level of statistical significance.  This Table shows only inflammatory and immune-related genes and therefore some genes in the top 50 in Table 1 do not appear here. This is now indicated in the legend.
As for Table 1, the level of significance should be indicated.
P<0.05 added to legend.
The Table could be made it easier to read if the "up-regulated (acute/control)" genes were shown below the "up-regulated (chronic/control)", rather than in adjacent columns, or were clearly separated using a thicker vertical line.
This change would make a very long 2 column table. We prefer the helpful suggestion that the columns should be separated which we have done with a narrow blank column In addition, it is confusing to have a column "chronic/control" under the "up-regulated (acute/control)" list. This seems to have been done to accommodate S100a rather than inserting it under each comparison.
Unfortunately in the editorial process of preparing the pdf from the submitted Table  spreadsheets some of the down-regulated genes have been sliced off and put incorrectly under the up-regulated categories. We are puzzled by this as the proof we received to check was correct. I have discussed this with the Editorial Office who have indicated that they will make sure this does not occur in the next version.
Please check carefully that the difference indicated in different Tables are the same (e.g. S100a8 has a FC 2.25 in Table1 and FC 2.26 in Table 2).
This was due to a difference in rounding, which has now been corrected.
Il1b is the notation used in the gene data base ncbi.nim.nih.gov, we would prefer to retain this notation in tables.
A pie chart of the inflammatory genes to complement Table 2 and Fig. 3 would be useful.
We generally find that pie charts are not helpful and would prefer not to make this addition.
P17, left column, top and Fig. 4. There is clearly variability, but to give numbers of fetuses where Il1ß levels could be detected over total numbers assayed for all groups would be more accurate and informative (e.g. acute 2/4, chronic low 7/16 and chronic high 10/19) than including these number only for the chronic high group, which appears to be wrongly given as 19/39, while the number of fetuses indicated for this group in Fig. 5 legend is 19.
We think that the Reviewer is probably referring to Fig 5. We agree that the way of representing these data that the Reviewer has suggested is clearer. This has been incorporated into the text (bottom P30). Figure 5 has been modified to make it clearer that values were obtained from 4 control fetuses. The legend has been re-written to make it clearer how many dams and fetuses were involved in this part of the study.
Figs 6 and 7 do not include error bars and no statistical analysis of these data seems to have been performed. It should be clearly indicated whether there was no statistical difference among groups at any time point studied.
Each point is a single fetus. The n values represent the number of fetuses in each treatment group. The legend has been rewritten to explain this more clearly.
P20, left column, top. The statement: "AFP levels were higher in all treated dams compared to an un-treated control." should be revised, as Fig. 8 shows an AFP increase only in chronically treated dams.
This has been revised to state that there was an increase in dams' AFP only in the chronically treated animals.
It is a pity that the number of dams is too small to assess the significance of this observation and that no housekeeping protein was used to normalize AFP expression. If B is a densitometry of the gel in A, where according to the western blot labelling and the legend there is only 1 control for both dam and fetus, why are there 2 samples indicated in the controls in the charts? Given the low sample numbers and variability, particularly in fetal AFP levels, expressing the data as ratio is not appropriate.
We agree that it is a pity that the numbers were very small, but we were constrained by the effects of being shut out of our laboratories for several months because of the coronavirus emergency.
In general the only way to obtain accurate AFP values is to measure the actual concentrations of the protein. We are very aware that Western blots are only semi-quantitative at best. We attempted to make the gels from which we took measurements as comparable as possible within each age group by using similar volumes of plasma (or diluted sample). The concentrations of plasma proteins vary between different animals and are not related to each other therefore using albumin as a reference protein would not provide more clarity. We thank the reviewer for drawing our attention to the discrepancy in control adult numbers in the western blots (A) and in the densitometry readings (B). This has been corrected.

Of note.
The paper is very complete. They demonstrate both acute and chronic changes to the placenta transcriptome with strong statistical relevance. Interestingly the chronic and acute genetic changes have few if any overlapping genes suggesting that long term toxicologic homeostasis may have very different effects to fetal development than single dosing. Thus, as noted by the authors, the use for the control of acute inflammatory responses may be warranted, but chronic ingestion any substance should be view with caution when the developing fetus is concerned.

Conclusion.
This paper is well conceived, clearly written and expertly interpreted. Safety profiles of drugs are in flux and whether vertebrate homeostatic metabolic responses to drug exposure, acute or chronic, is truly benign is an open question. These authors clearly demonstrate, by the discovery of soluble protein changes in dosing of paracetamol, that there is more to learn about drug toxicology, in particular at the chemoprotective interfaces of the body, in this case the placenta and BBB. While the consequences of these proteomic changes are unclear they are corroborated by profound compensations in the transcriptional profiles of the placenta. Interestingly, the placental barrier does the lion share of compensation as the BBB of pups is nearly unchanged. This is a reassuring finding for the developing brain, but leaves many unanswered questions about how the fetus may affect maternal physiology (as noted by the authors).
The implications of this study are profound and not only for the use of paracetamol. In this paper they describe a road map for the study of all drugs that could have maternal fetal interactions and provide the physiologic and genomic insights to back up their assertions. Indeed their proposed experiments in pregnant women to follow up on their findings would be very important to the management of pregnancy and to the field of maternal/fetal physiology as a whole.
I love this paper. BRAVO!

Major Issues
None.
Minor Issues. None.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound? Yes

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate?