Oxytocin (OT) is a 9-amino-acid neurohormone that is universally present in mammals and plays a key role in regulating a broad variety of social behaviors ranging from social memory and affiliation (sexual and parental behaviors, attachment, pair bonding) to aggression (“mate-guarding,” maternal aggression) (for reviews, see Lim and Young, 2006; Lee et al., 2009; Ziegler and Crockford, 2017). Emotional behaviors, such as anxiety-related behaviors and stress coping are also modulated by OT (Uvnäs-Moberg, 1998; Amico et al., 2008; Neumann, 2008). In particular, OT has been described as being involved in positive social interactions in intra-species interactions (Grewen et al., 2005; Baumgartner et al., 2008; Ditzen et al., 2009; Dunbar, 2010; Smith and Wang, 2012; Romero et al., 2015; Numan and Young, 2016) and inter-species interactions, such as human-animal interactions (Odendaal and Meintjes, 2003; Beetz et al., 2012; Rault, 2012, 2016; Romero et al., 2014). Some authors even suggest a role of OT in enhancing well-being (Uvnäs-Moberg and Petersson, 2005; Ishak et al., 2011).

Consequently, there is a growing interest in investigating and explaining the role of OT in behavioral neuroscience, which has led to an increasing number of studies on the topic. Some of these studies have focused on behavioral effects in different contexts of OT administration. Routes for OT administration are variable: of note, the intranasal administration of OT has recently solicited considerable interest because of its non-invasive nature, its ease of use, and its direct access to the brain, circumventing the blood-brain barrier and the rapid degradation of OT in blood (Veening and Olivier, 2013; for reviews, see Olff et al., 2013). Importantly, Lee et al. (2017) developed a method based on spectrometry to distinguish endogenous and exogenous OT that showed that intravenously or intranasally administrated OT were able to reach the brain in a similar way. Another approach investigating the behavioral/emotional effects of OT is based on correlational studies measuring OT levels in biological samples, such as excretory fluids (urine, saliva), blood or cerebrospinal fluid (Alves et al., 2015). However, the assessment and interpretation of urine and saliva measurements provide inconsistent findings and require further investigation, notably using efficient assay methods (Horvat-Gordon et al., 2005; Anderson, 2006; Carter et al., 2007; Young and Anderson, 2010). Conversely, plasma and CSF samples, despite being more invasive (though to a lesser extent for plasma), displayed interesting results with established correlations between OT concentrations in CSF and plasma samples in several species (Neumann et al., 2013; Carson et al., 2014; Dal Monte et al., 2014). Of note, Dal Monte et al. (2014) and Freeman et al. (2016) obtained contradictory results regarding the correlation of OT concentration in CSF and plasma in the same species Macaca mulatta, which can be explained by looking more closely at the measurement methods used: Dal Monte et al. (2014) used extracted plasma to assay OT and found correlation between OT levels in plasma and CSF while (Freeman et al., 2016) used unextracted plasma and did not find any correlation.

Indeed, there are several methodological issues regarding OT measurement reliability since the methods used to analyze plasma OT greatly vary in published reports that investigate the role of OT in social behavior, leading to some inconsistent results. The assay methods described in these studies vary widely and may or may not apply plasma extraction (with different types of extraction protocols), as previously highlighted (Christensen et al., 2014; Robinson et al., 2014). Historically, radioimmunoassay (RIA) procedures were used, usually involving prior extraction procedures (Kendrick et al., 1991; Gilbert et al., 1996; Wotjak et al., 1998; Cool and Debrosse, 2003; Handler et al., 2003). Little by little, enzyme immunoassays became more prevalent than RIA due to their safety and ease of use, and for unknown reasons, extraction steps were also gradually abandoned by some researchers as highlighted by Dal Monte et al. (2014), McCullough et al. (2013), and Nave et al. (2015). But, since the beginning of enzyme immunoassay (EIA) use to assay OT in plasma, some authors have added methodological caveats after having firmly demonstrated that extraction was crucial in order to remove interfering immunoreactive molecules and obtain consistent OT concentrations with the first studies involving RIA (Szeto et al., 2011; Robinson et al., 2014). Notably, in an elegant study, Szeto et al. (2011) proved the necessity of performing extraction both in RIA and EIA to eliminate compounds that artificially increase the apparent plasma OT levels. Indeed, they demonstrated that the linearity and the accuracy of measurements was only verified in extracted samples and that there was no correlation between OT values in extracted samples and OT values in unextracted samples. Additionally, using HPLC characterization of immunoreactive fractions, they showed that some non-OT immunoreactive products (with molecular masses larger than that of OT) persist even in the extracted samples. The authors suggest that these products could come from OT degradation. The critical importance of extracting plasma to assay OT in order to obtain reliable results was again underlined in various reviews (McCullough et al., 2013; Leng and Ludwig, 2016), and even by the EIA manufacturer (ENZO Life Sciences, 2015).

Furthermore, immunoassays require validation as highlighted and defined by Andreasson et al. (2015), DeSilva et al. (2012), and Kelley and DeSilva (2007). Specifically, the European Medicines Agency (EMA) set up guidelines for bioanalytical validation in 2011 (European Medicines Agency, 2011) to facilitate the adoption of these good analytical practices. In human pharmaceutical research, this approach is now standard. However, this is not necessarily the case when it comes to animal science. In addition, some immunoassay kits designed for use with human samples are directly used with animal samples without prior validation (Young and Anderson, 2010; McCullough et al., 2013). This can be problematic since basal levels of analyte may differ between species and may be subject to matrix effects due to the variable presence of interfering substances in biological fluids. It is then important to validate methods of measurement, even pre-existing or commercial methods, when changing the sample's origin and/or nature to obtain reliable results (Kelley and DeSilva, 2007; European Medicines Agency, 2011). The procedure for validating the method can be adapted according to the application and the intended use of the measured analyte (Lee et al., 2006; Valentin et al., 2011).

The goal of this study is to provide a detailed analytical validation of a reliable and universal method for assaying OT in plasma samples from various domestic animal species. Indeed, in some domestic species, such as pets and farm animals, plasma is the biological fluid most readily available, unlike CSF (which is assessed more easily in studies involving lab animals like mice or rats). The panel of domestic species was chosen to meet the needs that may exist in behavioral/emotional studies involving OT, notably on human-animal interactions (Coulon et al., 2012; Rault et al., 2013; Kis et al., 2015; Oliva et al., 2015; Thielke and Udell, 2017), maternal behavior (Castrén et al., 1993; Hernandez et al., 2002; Poindron et al., 2007; Ishii et al., 2008), or stress coping (Yayou et al., 2010; Sutherland and Tops, 2014), where OT can be assessed as a biomarker for a particular behavioral or emotional state. For the reasons stated earlier, we decided to perform an analytical validation based on the EMA guidelines for OT measurement using extracted plasma samples followed by EIA.

In this study, we assessed the suitability of an EIA kit (from Enzolifesciences) for accurately and precisely assaying OT concentrations in solid-phase extracted plasmas from seven domestic animal species. As results obtained in at least 67% of the tested VS were satisfactory regarding the acceptance criteria established in agreement with the EMA Guidelines and the Fit-For-Purpose principle (Lee et al., 2006) for within-run precision and extraction efficiency, we can conclude that the use of this method has been validated for the measurement of plasma OT levels in dog, cat, horse, cow, sheep, goat and pig samples. Furthermore, 100% of VS fulfilled all acceptance criteria in cats, sheep, goats, and pigs. Between-run precisions were also satisfactory and yet highlighted substantial variations in the measurements near the LLOQ values. This was observed in almost all studied species and consequently the authors recommend cautiously interpreting the results obtained in these lowest ranges of concentrations, especially when several runs are needed. For this reason, and because of the sensitivity limit of the kit at 15 pg/ml, we decided to perform sample concentration during the extraction steps in almost all species, except for dogs and cats for which it was not necessary to do so. Indeed, in these species, the basal OT values found in this study were above the kit's LOD. For the other species, assay attempts were made without concentrating their plasma, but were unsuccessful as we found out afterwards that their basal peripheral levels of OT were below the kit LOD. Sample concentration was thus necessary to reach OT values in the dynamic range of the EIA kit and concentration factors were adapted to each species after several analyses. Nevertheless, sample concentration required higher amounts of plasma, and thus a greater volume of blood, which must be drawn without jeopardizing animal health. This has to be taken into account from an ethical point of view and considering the 3Rs rule (Franco and Olsson, 2014). In the case of domestic animals enrolled in this study, they were of sufficient size to safely supply the volume of blood required.

Here, the authors have chosen to perform a partial validation since the kit was of a commercial nature and already validated for use in human plasma. Indeed, the EMA Guidelines state that a full validation is not required when a change in biological matrix occurs, which is the case in this study, with plasma originating from domestic animals. A partial validation can range from as little as the determination of the within-run precision and accuracy to a nearly full validation. In our partial validation, we established the precision profile of OT measurements in each studied species and assessed the extraction efficiency (reflecting the global procedure accuracy as explained before), within-run and between-run precisions. Importantly, despite precision and extraction efficiency yielded satisfactory results, we were not able to assess matrix effects. Even if the SPE method and the use of assay buffer as diluent to reconstitute samples are expected to remove interfering substances and minimize the risk of matrix effects (Tate and Ward, 2004), we cannot be sure that all potentially interfering substances are indeed removed, especially after concentration of the plasma samples. So, results should still be interpreted with caution, as perhaps discrepancies might be found because of putative persisting matrix effects in some cases. Despite their importance, parallelism studies between the calibration standard curve and serially diluted study samples could not be performed due to very low basal levels of OT in each species, which made it impossible to obtain the needed serial dilutions for this kind of study, even after concentrating the samples many times. As already stated, the sample concentration required a large amount of plasma to be taken, which was a real constraint depending on the studied species, notably in the smallest animals. In addition, handling large volumes of plasma is a source of increasing errors and practical complications entailed with the use of HyperSep C18 columns, which can clog. So further increasing the concentration factors to artificially elevate the sample OT concentrations in order to carry out parallelism studies was not a desirable option, considering that it would also have alienated the aim from using real samples reflecting the actual study populations, as recommended by Valentin et al. (2011). Finally, a parallelism study can also provide doubtful evidence, as highlighted by McCullough et al. (2013) about the results obtained without extraction steps in Kramer et al. (2004), which were subsequently questioned by Szeto et al. (2011). Besides, because of the prior extraction steps, we did not purposely include lipemic or hemolytic samples in our study. Indeed, in a series of experiments, we observed that extraction steps were effective enough to prevent any discrepancies in the measurements. In the present study, dilution was never required, and samples from 5/7 species even had to be concentrated. However, to provide data on the resemblance between the endogenous analyte and the standard, we performed dilutional linearity tests on dogs and cats' samples, where no concentration was required. The linearity under dilution was ascertained in both samples origin, throughout the linear range of the kit. Despite the differences of biological matrix and the putative outcomes, it is also interesting to note that the dilutional linearity was demonstrated by the manufacturer in human serum (ENZO Life Sciences, 2015). As stated by Lee et al. (2006), the proven dilutional linearity could support parallelism.

Furthermore, the fit-for-purpose principle states that “assay validation should be tailored to meet the intended purpose of the biomarker study, with a level of rigor commensurate with the intended use of data” (Lee et al., 2006). In the present context of research on the link between OT and multiple behaviors or emotional status in domestic animals, OT can be considered as an exploratory biomarker. This is why the analytical validation performed here comprised 3 VS levels, as well as 3 independent runs with 3 replicates per run for precision and accuracy assessment, as supported by the literature (Lee et al., 2006; Valentin et al., 2011). In addition, we chose triplicate analysis over the “classical” duplicate analysis since we observed a significant within-run variability in some cases (DeSilva et al., 2012). In particular, we noticed that insoluble material after sample reconstitution in assay buffer could heavily influence the calculated concentrations, which may be falsely overestimated due to a decrease in OD in the wells where insoluble material has been observed throughout the EIA procedure. The presence of insoluble material was especially noticeable in dog plasma samples and proved very difficult to remove. To resolve this issue, we added a step to the procedure for all species; samples were spun briefly after the reconstitution in assay buffer and were then carefully pipetted into the plate wells.

Interestingly, the precision profiles of neat samples showed that the basal plasma levels of OT observed in every species involved in this study were coherent with those previously described in the literature. For dogs, when samples were extracted, OT mean concentrations using RIA were found to be approximately 63 pg/ml in bitches in anestrus (Olsson et al., 2003) and 45 ± 40 or 65 ± 82 pg/ml (in two different breed populations) using the same EIA method (Hollinshead et al., 2010). There are however marked differences with dogs plasma OT concentration measurements from studies in which the plasma was not extracted (Handlin et al., 2011; Rehn et al., 2013); this discrepancy is to be expected since such discrepancies have already been extensively documented in the literature from human samples (Szeto et al., 2011; McCullough et al., 2013; Christensen et al., 2014) where higher OT concentrations in unextracted plasmas have been demonstrated to be due to the presence of non-OT immunoreactive products. In cats, the authors found only one study dealing with plasma OT concentrations in the species that reported a basal level between 20 and 30 pg/ml in solid-phase extracted samples and assayed with “in-house” EIA (Fieni et al., 2006). In three studies (Ginther and Beg, 2009, 2011, 2012) using horse samples and a very similar method of extraction and assay comprising a concentration factor of 13, Ginther and Beg found OT concentrations between 10 and 35 pg/ml, which are close to the values we found once the concentration factors have been applied. In cows, studies using RIA methods (Mačuhová et al., 2004; Belo and Bruckmaier, 2010) found basal levels of plasma OT between 1–4 and 3–6 pg/ml, which are similar to our results. Again, predictably, authors (Sutherland et al., 2012) using EIA methods with unextracted cow plasma exhibited OT concentrations 100-times higher than in the present study and studies using RIA. In sheep, basal plasma OT concentration was measured using RIA in the 5–10 pg/ml interval in extracted samples from ewes (Dwyer et al., 2004). In goats, reports using extracted samples and RIA or “in-house” EIA found basal plasma OT concentrations around 10 pg/ml (Payne and Cooke, 1998) or respectively in the interval of 14–47 pg/ml (Hernandez et al., 2002; Olsson and Högberg, 2009). Finally, basal plasma OT levels in pigs were described around 5–20 pg/ml in experiments using RIA following extraction (Castrén et al., 1993; Gilbert et al., 1996). This thorough comparison of our findings with the literature in every species supports the use of the method described and analytically validated here as a suitable means of measuring relevant OT measurements in the plasma of the domestic animal species involved in the study.

Of note, basal plasma OT levels among the seven studied species varied greatly. To the best of our knowledge, the reason for this species differences is not known, although our main assumption relates to the animals' size and the total volume of blood in each species: indeed, we experimentally noticed that the smaller the species is, the higher the basal levels of OT are. This would be interesting to further investigate it. Likewise, the question of sex differences in OT plasma levels is noteworthy when considering the increasing number of reports stating the sex importance in relation to the OT system (Neumann, 2008; Bakermans-Kranenburg and van IJzendoorn, 2013; Crockford et al., 2014; Alves et al., 2015), notably through the influence of steroid hormones, such as estrogen (Murakami et al., 2011; Acevedo-Rodriguez et al., 2015). Unfortunately, in this study, we did not specifically record the sexes of the enrolled animals as our purpose was to describe and validate a method which could be useful for numerous domestic species, regardless of the sex. Because the method has been especially validated on individuals from both sexes, we indeed think the method presented here could be helpful in studies investigating the sex effect on OT plasma levels. However, our recommendation would be to assay the samples from different animals' sexes on the same plate because the possible between-run variability found here may not permit to highlight small putative differences between sexes.

The procedure described in this study for assaying OT is mainly based on the procedure assessed by Szeto et al. (2011) for human plasma samples. These authors suggested that even extracted plasma contained non-OT immunoreactive products that could come from OT degradation. Thus, when we refer to OT, it should be understood as “oxytocin-like immunoreactive products.” Recently, some authors (Brandtzaeg et al., 2016) have described a new procedure to assay the total fraction of OT in plasma, in contrast with all previously published methods, including the one presented here, which aim to measure the free fraction of OT in plasma. They found that OT could bind to plasma proteins and thus be precipitated with them during the extraction procedure. In this regard, non-extracted protocols of OT assay may have partial immunoreactivity with bound OT, that could explain the method discrepancies. Treating the plasmas by reduction/alkylation could release OT molecules from plasma proteins, hence allowing higher amounts of OT to be measured, i.e., the total plasma OT fraction, notably in dog plasma. This could eliminate some drawbacks of the present method, such as the need to considerably increase the concentration of domestic animal samples, which requires drawing substantial amounts of blood. However, the biological activity of the bound fraction needs further clarification. Likewise, the interest of OT total fraction as a behavioral/emotional biomarker remains to be investigated.

Method validation is a fastidious but critical step to providing reliable measurements and supporting strong scientific findings. The field of OT research is rapidly expanding. In human science, elegant advancements have been made to improve OT measurements and to establish standard, repeatable, reliable and validated methods to achieve them. Applying them to plasma samples of seven domestic animal species, the authors hope this study will help to provide reliable tools for measuring peripheral OT and supporting future studies on domestic animals in OT research fields.

CB, AC, and PP wrote the manuscript and conceived the study. CB and CC analyzed the data, conceived, designed and performed the experiments. EC performed and supervised the blood sampling procedures in all species.