Preprint: A robust peptidomics mass spectrometry platform for measuring oxytocin in plasma and serum

Current approaches to measuring the cyclic peptide oxytocin in plasma/serum are associated with poor selectivity and/or inadequate sensitivity. We here describe a high performance nano liquid chromatography-mass spectrometry platform for measuring OT in human plasma/serum. The platform is extremely robust, allowing laborious sample clean-up steps to be omitted. OT binds strongly to plasma proteins, but a reduction/alkylation procedure breaks this bond, allowing ample detection of total OT. The method showed excellent quantitation properties, and was used to determine total OT levels to 0.5-1.2 ng/mL (evaluated with human plasma and cord serum). The method is compatible with accessible mass spectrometry instrumentation, finally allowing selective and easily comparable oxytocin measurements.


Introduction 1
The neuropeptide oxytocin (OT) is a facilitator of childbirth and breastfeeding, and can activate 2 maternal behavior 1 and partner preference 2 in animal models. In humans, OT levels have been 3 related to e.g. autism 3 , and schizophrenia 4 . Several studies have reported a coordinated release of 4 central and peripheral OT 5,6 and that peripheral levels can be a low--invasive indicator of central state 5 7 . However, the brain/blood OT relation is a disputed topic 8 casting doubt on the biomarker--ness of 6 peripheral OT. A key source of skepticism is the absence of satisfactory analytical methodology of OT 7 measurements 9 . Nearly without exception, enzyme--linked immunosorbent assays (ELISA) and 8 radioimmunoassays (RIA) are used to monitor OT in blood and other biofluids. These methods have in 9 recent years been severely criticized due to poor selectivity 8,9 . An alternative to ELISA/RIA is mass 10 spectrometry (MS). The MS instrument allows unambiguous identification/quantification of e.g. 11 peptides, by first recording the molecular mass of a compound (single MS), and then creating a 12 molecular "fingerprint" by fragmenting the compound to smaller parts (MS/MS). Separating 13 compounds in a mixture (e.g. plasma) prior to MS detection further strengthens identification and 14 sensitivity. Peptides are typically separated using liquid chromatography (LC). LC--MS is an invaluable 15 tool in virtually all areas of biomedical analysis. A notable exception is however OT measurement; the 16 few published methods for LC--MS measurements of plasma OT 10,11 provide unsatisfactory sensitivity 17 and varying results, and are therefore difficult to put to practical use. We here set out to develop a 18 robust and sensitive method for quantification of OT in blood, as a remedy for the under--par LC--MS 19 and ELISA/RIA performance regarding OT analysis. We here "borrow" tools from mass spectrometry 20 based proteomics, namely i) nanoLC--MS (a particularly sensitive variant of LC--MS 12 ) featuring on--line 21 sample extraction 13 , and ii) a reduction/alkylation step 14 , allowing vastly increased OT extraction.

Enabling nanoLC--MS for robust and simple plasma analysis 2
NanoLC--MS is exceptionally sensitive and selective instrumentation for identifying and measuring e.g. 3 peptides 12 , and is commonplace in proteomics facilities. However, nanoLC--MS is rarely used for large--4 scale blood/serum/plasma sample analysis, in part due to its limited robustness (i.e. clogs easily if 5 extensive sample preparation is not undertaken). This weakness was overcome by implementing an 6 automated filtration/filter back--flush (AFFL) unit 15 to the nanoLC--MS system, allowing robust plasma 7 analysis. Details are described below. 8 In preliminary experiments with a standard nanoLC--MS set--up (i.e. trap column for extraction + 9 separation column), injecting protein precipitated pooled human plasma clogged the column(s) (see 10 Sensitive and stable detection of plasma OT following a reduction/alkylation step 14 We find that OT strongly binds to plasma proteins, which can seriously affect the sensitivity/measuring 15 accuracy in biomarker studies. However, performing a reduction and alkylation step liberates OT from 16 plasma proteins, allowing ample sensitivity and precise quantification of endogenous (total) OT. Details 1 are described below. 2 Initially, samples contained 50 mM ZnCl 2 (10 mM aspartate buffer, pH 4.5) to stabilize OT via chelation 3 16 prior to subsequent sample preparation (e.g. removing proteins via protein precipitation (PPT)). 4 However, adding ZnCl 2 to plasma samples resulted in noisy signals and pressure build--up, likely due to 5 on--column precipitation of salts and/or proteins. Acetonitrile based PPT (without the presence of 6 chelating agents) was associated with an unassuring recovery profile (OT recovery dropped and leveled 7 off after 40 minutes ( Figure SM 1)). OT was stable in the solvents used during and after PPT ( Figure SM  8 2), and did not absorb to tubes and vials. It was considered unlikely that the main metabolizing enzyme 9 for OT in plasma, cystinyl aminopeptidase/oxytocinase 17 was degrading OT in these conditions, as this 10 enzyme is rather large (subject to PPT), and blood from non--pregnant individuals was used. Therefore, 11 we speculated that the recovery profile depicted a slow binding to protein remains. To further assess 12 the issue of OT protein binding, pooled human plasma was spiked with oxytocin, and was stored on 13 the laboratory bench up to 8 h before PPT; recovery of the spiked OT linearly deteriorated as function 14 of time before the PPT step (Figure SM 3), once again suggesting a slow and strong protein binding 15 after spiking. Furthermore, OT spiked to plasma had very poor filtrate recovery using size separation 16 with centrifugal filters, again implying strong protein binding. 17 We hypothesized that strong protein binding was preventing detection of endogenous OT ((<pg/mL 18 levels, Figure SM

6
OT was determined in pooled plasma and human cord serum, obtained from commercial sources: The 7 concentration of oxytocin in pooled human plasma from Sigma Aldrich and Innovative Research was 8 0.5 ng/mL and 0.7 ng/mL, respectively. For pooled human cord serum (Innovative Research) the OT 9 concentration was expectedly higher 18 , 1.2 ng/mL (Figure 3a). Oxytocin plasma levels were, as 10 expected, higher after nasal intake of OT (Figure 3b). However, the fold--change was very dependent 11 on the individual. For instance, person 2 (who described him/herself as highly anxious prior to sample 12 collection) had a markedly different OT plasma profile before/after intranasal administration. Our 13 results confirm the common assumption that OT levels can significantly vary between individuals 19 14 (Identification/quantification of OT was based on using external standards, a deuterated internal  Discussion 8 A reduction and alkylation step was key in "liberating" oxytocin from plasma proteins, allowing ample 9 detection of endogenous high pg--ng/mL amounts in human plasma. Tight plasma binding is not 10 uncommon with biomarkers 20 . The OT levels observed in this study are remarkably higher compared 11 to that obtained with an off--line extraction step (low pg/mL levels) 21 . With extraction, the vast 12 majority of OT is discarded with plasma proteins, leaving only a minute free amount of OT left to be 13 measured. Measuring only the free fraction, as currently recommended (ref leng, mccullogh) can be a 14 confounding factor, since the free OT concentration can be drastically changed by factors such as age, 15 morbidity, or by compounds that displace OT from proteins 22 . This is especially the case if the marker 16 is heavily bound 22 , as we find with OT. Indeed, even using MS large variations are observed when 17 measuring the free fraction of OT; a third of the human samples analyzed by Zhang et al did not 18 contain detectable levels of OT 10 . We have also registered such inconsistencies with our own 19 "neurotransmitter--omics" MS platform 11 . In addition, free OT levels varied 6--fold within a 1 homogenous group of rats 10 . As shown in Figure 3, when all circulating OT is measured using our 2 method there differences between individuals are already pronounced individuals (but not unusually 3 large compared to much of the metabolome). Such individual differences are thought to be highly 4 informative 19,23 ; additional confounding factors will undoubtedly make correlations less clear. Based 5 on this reasoning, total OT is better suited as a biomarker than only the free fraction of OT. 6 Considering the growing concerns of antibody--based assays (both RIA and ELISA fall in to this category) 7 regarding selectivity and antibody kit reproducibility 24 , LC--MS is a natural choice for OT measurements 8 due to its excellent selectivity. The robust and highly automated AFFL--nanoLC--MS approach has 9 attractive quantification traits, and can be simply implemented in e.g. proteomics facilities (common in 10 e.g. many larger universities/hospitals). As the instrumentation is compatible with salty solutions, 11 AFFL--nanoLC--MS can be used for urine and cerebral spinal fluid measurements as well (protein binding 12 can also occur in these matrices). Other LC--MS systems can be employed, e.g. UPLC--MS systems used 13 for drug measurements or metabolomics, but these may require off--line filtration/extraction steps. 14 15

Nasal spray experiment 6
Three healthy volunteers, one female and two males were asked to apply two puffs of OT nasal spray 7 (6.7 µg OT/puff, Syntocinon® from Sigma--Tau Pharmaceuticals, inc., Gaithersburg, MD, USA) in each 8 nostril. The subjects were asked to spray close to the respiratory region, where there has previously 9 been shown best absorption 25 . Two blood samples were drawn from each participant; one 5 min and 10 another 20 min after the puffs of OT nasal spray were applied. Plasma was made and the samples were 11 analyzed the same day (n=4). MS--software was used in the AFFL system. See Figure  1 for plumbing of the AFFL system. A Hitachi L--20 7100 HPLC pump (Chiyoda, Tokyo, Japan) in isocratic mode was used to back--flush the filter in the AFFL 21 system with type 1 water. In position 1 (Figure  1  A), the sample passed through a stainless steel filter 22 (1 µm porosity, 1/16" screen, VICI) onto a 100 µm ID x 50 mm silica monolithic C18 SPE manufactured 23 as described in 26 (similar to Chromolith® CapRod C18 capillary columns from Merck Millipore). In 1 position 2 (Figure 1 B), two processes happened simultaneously; the filter is back--flushed, while 2 oxytocin is back--flushed from the SPE column onto a 100 µm ID x 150 mm silica monolithic C18 3 separation column manufactured as described in 26 (similar to Chromolith® CapRod C18 capillary 4 columns from Merck Millipore). A steel emitter, 30 µm ID x 40 mm, from Thermo Scientific, was 5 connected to the end of the separation column by a 1/16" standard steel internal union from VICI. A 6 nanospray Flex TM ion source (nanoESI) coupled to a Quantiva TM triple quadrupole mass spectrometer 7 from Thermo Scientific was used for detection of oxytocin in full MS--and tandem MS--mode (MS/MS). 8 9

Liquid chromatography and mass spectrometry parameters 10
The 20 min gradient program was composed as follows: 20 %B isocratic elution for 14 min, followed by 11 an increase from 20 to 90 % B in 2 min before isocratic elution at 90 % B for 4 min. used with 3 mTorr collision--induced dissociation (CID) gas. Argon was used as collision gas. In addition, 1 25 V source fragmentation energy was used together with 3 secs chrom filter. 2 3

Data analysis and interpretation 4
Data analysis and interpretation were done using Xcalibur TM software version 3.0 from Thermo 5 Scientific. 6 7

Ethical statement 8
All subjects gave written informed consent, and the blood collection was approved by the Regional 9 Ethics Committee (2011/1337/REK S--OE D). All methods were carried out in accordance with the 10 approved guidelines and regulations. 11 12 13