**3. Discussion**

In addition to widespread endothelial dysfunction, alterations to maternal, placental and fetal metabolism are hallmarks of PE [18]. In this study, we explored third trimester placental creatine metabolism in early-onset PE and gestation-matched normotensive pregnancies. Our findings confirm the recent metabolomic reports of altered creatine homeostasis with PE [9]; moreover, we identified significant changes in expression of genes associated with the synthesis and transport of creatine within the placenta. Our main finding was that total creatine content of the third trimester PE placentae was significantly higher than controls. Similar adaptations have recently been described in placental insu fficiency leading to fetal growth restriction (FGR) in the absence of PE [17] and in pregnancies that occurred at high altitude [19]. To the best of our knowledge, these studies are the first to describe increased creatine content in human tissue exposed to perturbations in oxygen delivery, in vivo.

There are three potential mechanisms for changes in placental creatine content between normotensive and PE placentae. These include changes to creatine to creatinine degradation rates, increased endogenous synthesis and/or increased cellular uptake of creatine from the circulation. It is unlikely that changes to the rate of creatine degradation to creatinine would have led to di fferences in placental creatine content between cohorts, as this is a spontaneous, non-enzymatically driven process that occurs at a near constant rate of ~1.7% of total body creatine content per day [20]. This leaves changes to endogenous placental creatine synthesis and/or rates of cellular up-take as potential mechanisms for increasing placental creatine content with PE. As cellular metabolism is in a constant state of flux, it is di fficult to ascertain exactly which of these processes (de novo synthesis or cellular up-take, or both) resulted in the increased total creatine content of the PE placenta in this retrospective study. However, we make the following observations.

PE up-regulated placental gene expression of the creatine synthesizing enzymes *GATM* and *GAMT*. Modulation of *de novo* creatine synthesis is usually discussed in the context of AGAT (i.e., *GATM* mRNA expression and/or AGAT activity) as this is thought to be the rate-limiting step of creatine biosynthesis [21]. Mediators of AGAT expression and activity include thyroid hormone, growth hormone, and circulating levels of arginine, citrulline and creatine [22,23]. The influence of arginine, citrulline and creatine on the rate of endogenous creatine synthesis has mainly been described as a down-regulation of *GATM* and AGAT when these metabolites are in excess, which is the opposite to what has been observed in the PE placentae [22]. On the contrary, adequate concentrations of thyroid and growth hormone are required to maintain AGAT activity, with studies in thyroidectomy or hypophysectomy rats observing decreased AGAT renal activity [24]. Whether increases in thyroid or growth hormone above basal levels can increase AGAT activity is unknown. There is the potential that increased placental growth factor production by the PE placenta may be influencing these transcriptional pathways, and maintaining the placental *GATM* mRNA expression with advancing gestational age and birth weight observed in the current study, a finding that was absent in normotensive controls [25].

Despite increases in gene expression of both *GATM* and *GAMT* and the increased total creatine content, there were no detectable changes in AGAT or GAMT protein levels, indicating a mismatch in transcription and abundance of the enzymatic machinery of creatine synthesis in the PE placentae. It may be that placental endoplasmic reticulum stress reported with PE may hinder increased production of AGAT and GAMT enzymes [26]. There is also the potential of an increased rate of *de novo* creatine synthesis by increased activity of the AGAT and GAMT enzymes in the absence of changes to overall protein abundance. As no change in placental GAA content was observed between cohorts, we would contend that any changes in AGAT and GAMT activity would have been in equilibrium. We cannot rule out how histomorphological di fferences between control and PE placentae may have influenced measured AGAT and GAMT expression, as our assessments were conducted on tissue homogenates. An increase in terminal villi density, but an overall decline to villous surface area are characteristic of placental PE histology [27]. AGAT is expressed in stromal and endothelial cells of the fetal capillaries and GAMT in the syncytiotrophoblast of the fetal villi [10], so it is plausible that levels of AGAT and GAMT could be altered by changes to cell populations. However, one would anticipate that the known

histological changes within the PE placenta would lead to a down-regulation in expression, not the increases in mRNA expression and no change in protein levels observed in the current cohort.

The alternate route of increasing placental creatine content is by increased transport of creatine into the placental cells via the creatine transporter [28]. This could simply be the result of increased maternal circulating creatine levels with PE. While no studies have characterized changes in maternal creatine levels in PE cases, increased creatine in cord plasma has been reported [9]. The retrospective nature of the current study and lack of maternal blood sampling inhibited our ability to explore this hypothesis further. Assessment of maternal circulating concentrations of creatine should be a focus of future prospective cohort studies. The current study did find a 2-fold increase mRNA expression of the creatine transporter *SLC6A8* in PE placentae compared to controls. A potential mechanistic pathway for increased creatine transporter expression and activity in the placenta is via AMP-activated protein kinase (AMPK), a key regulator of cellular energy metabolism that has been implicated in the pathophysiology of PE [29]. In cardiomyocytes, activation of AMPK pathways has been shown to up-regulate creatine transporter expression and activity, ultimately increasing cellular creatine up-take [30]. This effect is thought to be indirect, as the creatine transporter does not contain substrate sites for AMPK mediated phosphorylation. Potential intermediates are co-activator peroxisome activated receptor γ coactivator-1 alpha (PGC-1α) and estrogen-related receptor α (ERRα), both known to be regulated by AMPK and induce the expression of genes involved in cellular bioenergetics, including mitogenesis [31]. Studies in skeletal muscle cells (L6 myotubes) have described direct interactions between PGC-1α/ERRα, increased creatine transporter mRNA expression and cellular creatine up-take [32]. As there have been mixed reports of increases and decreases to placental mitochondrial content and PGC-1α expression with PE [33,34], further exploration of this mechanistic pathway is required before conclusions about AMPK's role in driving changes to creatine transport and cellular up-take in PE placentae can be drawn. In vitro techniques should be employed for these analyses to control for changes in placental metabolism associated with tissue collection and processing times.

It is probable that placental mitochondrial dysfunction [8], and thus changes in AMPK expression, are also influencing how the PE placenta is utilizing the available creatine for ATP homeostasis. AMPK is a known regulator of the expression of the cytosolic isoform of creatine kinase *BBCK* (responsible for the hydrolysis of phosphocreatine and phosphorylation of ADP to form ATP) [35]. This finding is consistent with the increased *BBCK* mRNA expression observed in PE placentae in the current study. Activation of the HIF-2α hypoxia response element, which has been shown to be up-regulated in PE, may also contribute to changes in *BBCK* mRNA expression, as activation of HIF-2α enhances BBCK expression in apical intestinal epithelial cells with inflammatory bowel disease [36]. Again, there was no detectable difference in protein abundance in the current study. Thus, further interrogation of these mechanisms is required to elucidate the interactions between metabolic regulatory pathways and creatine metabolism in the PE placenta. Known variations in mitochondrial function and thus capacity to generate ATP between pre-term and term PE placentae should be considered within this context [37].

Interestingly, the mechanism (de novo synthesis and/or cellular uptake) for increasing placental creatine levels may be different between the PE and FGR pathologies. Placentae of growth restricted infants had lowered GAA levels and reduced mRNA expression of *GATM* and *GAMT*, compared to the PE placentae that showed no change in GAA content and an up-regulation in *GATM* and *GAMT* [17]. It is generally accepted that the level of placental insult is higher in PE than in FGR, producing a more substantial burden of placental pro-inflammatory and anti-angiogenic factors and their interactions on maternal endothelium in multiple-organ systems [38]. As maternal kidney injury is a primary manifestation of the PE syndrome, changes to maternal systemic creatine synthesis, mainly GAA production by AGAT activity in the renal proximal tubules, may alter maternal creatine homeostasis, placing a more significant burden on the placenta to maintain adequate creatine levels for placental and fetal requirements.

Increased de novo creatine synthesis, either by the placenta or maternal production by the kidney and liver could have further implications for the pathogenesis of PE. The methylation of GAA by GAMT produces creatine and S-Adenosylhomocysteine, contributing to around 40% of total homocysteine production under basal conditions [39]. Elevated maternal homocysteine levels have been identified in several PE cohort studies. As homocysteine can cause endothelial dysfunction, it is thought that hyperhomocysteinemia could contribute to maternal vascular dysfunction with PE [40,41]. Considering that a consequence of increased *de novo* creatine synthesis in response to PE could be the over-production of homocysteine, understanding these pathways may have implications for the managemen<sup>t</sup> of the maternal PE syndrome beyond placental bioenergetics.

There were several limitations to this study. Villous tissue was dissected from a central placental cotyledon only, and not randomly sampled from several sites across the placenta, as is now the recommendation for molecular studies [42]. The retrospective nature of this study also meant that some maternal demographic information that may have been relevant (i.e., ethnicity) was not available. Finally, the lack of an appropriate antibody inhibited our ability to quantify creatine transporter protein concentrations, via western blot [28].
