Pre-treatment of blood samples reveal normal blood hypocretin/orexin signal in narcolepsy type 1

Abstract The hypocretin/orexin system regulates arousal through central nervous system mechanisms and plays an important role in sleep, wakefulness and energy homeostasis. It is unclear whether hypocretin peptides are also present in blood due to difficulties in measuring reliable and reproducible levels of the peptides in blood samples. Lack of hypocretin signalling causes the sleep disorder narcolepsy type 1, and low concentration of cerebrospinal fluid hypocretin-1/orexin-A peptide is a hallmark of the disease. This measurement has high diagnostic value, but performing a lumbar puncture is not without discomfort and possible complications for the patient. A blood-based test to assess hypocretin-1 deficiency would therefore be of obvious benefit. We here demonstrate that heating plasma or serum samples to 65°C for 30 min at pH 8 significantly increases hypocretin-1 immunoreactivity enabling stable and reproducible measurement of hypocretin-1 in blood samples. Specificity of the signal was verified by high-performance liquid chromatography and by measuring blood samples from mice lacking hypocretin. Unspecific background signal in the assay was high. Using our method, we show that hypocretin-1 immunoreactivity in blood samples from narcolepsy type 1 patients does not differ from the levels detected in control samples. The data presented here suggest that hypocretin-1 is present in the blood stream in the low picograms per millilitres range and that peripheral hypocretin-1 concentrations are unchanged in narcolepsy type 1.


Introduction
Since the discovery of the hypocretin/orexin (HCRT) peptides in 1998, 1,2 much have been learned about their functions. They are implicated in a wide range of neuronal circuits, playing important roles in sleep and wake regulation, autonomic regulation, neuroendocrine and locomotor function, regulation of hormonal secretions, feeding behaviour and energy homeostasis as well as in the pathology of Narcolepsy. [3][4][5] The two hypocretin neuropeptides are derived from a common precursor protein (prepro-HCRT) which is proteolytically cleaved to form the 33-amino acid peptide hypocretin-1 (HCRT-1) (3.56 kDa) and the 28-amino acid peptide hypocretin-2 (HCRT-2) (2.94 kDa). 2 The peptides are produced by 60 000-70 000 neurons (in humans) located in the lateral hypothalamus, 6 from where the neurons project to most areas of the brain. Interestingly, a growing body of evidence points towards a direct regulatory effect of HCRT peptides, not only in the brain, but also in peripheral tissues. HCRT receptors have been found in a variety of tissues including kidney, adipose tissue, gastrointestinal tract, pancreas, adrenal gland and reproductive organs across different species. [7][8][9][10][11][12][13][14] The abundance and wide distribution of HCRT receptors in peripheral tissues suggests biological relevance, but for many tissues the function remains elusive. The source of peripheral HCRT is also unknown. Detection of prepro-HCRT mRNA and HCRT peptides in peripheral organs (stomach, ileum, colon, adrenal glands and pancreas) has been reported but the specificity of the signals have never been verified. 7,9,13,15 Common for these findings is that the presence of peptide has only been shown using antibody-based technologies, and the specificity of these antibodies have been questioned. 16 Peripheral HCRT might also derive from the central nervous system and function as a peptide hormone. It has indeed been demonstrated that HCRT-1 is highly lipophilic and can cross the blood brain barrier Graphical Abstract from blood 17 and that injections of HCRT-1 into the CSF of the cisterna magna can reach the blood. 18 The presence and possible transport of HCRT peptides in the blood stream has, however, been a topic of much debate. The controversy originates from difficulties in measuring reliable and reproducible levels of HCRT-1 or HCRT-2 in serum or plasma and lack of consistency and specificity of the methods/antibodies used. 19,20 Supplementary Table 1 summarizes findings from published studies claiming to measure HCRT-1 in blood (list not exhaustive). Despite applying similar methodology for detection and quantification of HCRT-1, markedly different HCRT-1 levels have been reported in both plasma and serum. To date no study of peripheral HCRT-1 levels has convincingly demonstrated that the signal observed was a genuine HCRT-1 signal.
Adding to the controversy are two conflicting papers describing measurements of plasma HCRT-1 levels in narcolepsy type 1 (NT1) patients. The sleep disorder NT1 is associated with a loss of HCRT production in the brain: low levels of HCRT-1 in the CSF is a highly specific and sensitive diagnostic marker for NT1. 21 To test if plasma HCRT-1 could be used as a biomarker, early attempts were made towards measuring HCRT-1 in blood samples from these patients and controls. While Dalal et al. 22 showed that plasma HCRT-1 levels were normal in narcoleptic patients Higuchi et al. 23 showed decreased plasma levels in narcolepsy. A later commentary questioned the specificity of the assays used in the two studies. 24 A reproducible and reliable assay that gives an authentic HCRT-1 signal from blood samples is still to be developed. Therefore, we aimed to develop and validate a simple method that allows for detection of HCRT-1 in blood. In addition, we evaluated if our assay would be useful as a diagnostic tool in NT1.

Outline of experiments
To meet our goals, we performed the following series of experiments: (i) Assay development, where we tested a whole range of sample pre-treatments to find the optimal conditions for HCRT-1 detection. (ii) Assay validation, where we verified the specificity of the HCRT-1 signal by using different antibodies, different methodologies and finally by using blood from animals lacking HCRT-1 in blood. (iii) HCRT-1 measurements in blood from two independent cohorts of NT1 patients and clinical controls.

Human samples for assay development
Blood was collected by the Department of Clinical Biochemistry at Rigshospitalet, Glostrup. All human samples were completely anonymized by pooling and by removing any personal references from the materials.
Blood samples were drawn into ethylenediaminetetraacetic acid (EDTA) or serum Vacutainer tubes (BD, Franklin, NJ, USA). EDTA samples were gently mixed and then immediately centrifuged at 2000 g for 15 min at 4 C. Serum samples were gently mixed, allowed to clot for 30 min and then centrifuged at 1500 g for 10 min at 4 C. Following clinical analysis, an estimate of 50 plasma or serum samples were pooled to obtain a plasma and serum pool, respectively. The time from collection to pick-up was 3-4 h and meanwhile the samples were kept at room temperature. After pooling, plasma and serum were aliquoted and stored at À80 C until use. For evaluation of assay interference by plasma anticoagulants and serum, blood from two individuals was drawn into citrate-, EDTA-, heparin-and serum Vacutainer tubes (BD). Plasma and serum were prepared as described above. Half of each sample were frozen for 30 min at À80 C prior to standard operating procedure (SOP) treatment, whereas the other half were SOP treated without freezing. Animals C57BL/6 male and female Hcrt-KO (hetero-and homozygous, n ¼ 10 and n ¼ 26), Hcrt-Atxn3 (n ¼ 10) and wildtype (WT, n ¼ 30) mice were used for these experiments. The Hcrt-Atxn3 mice model is a HCRT-specific neurodegeneration model. 25 Blood from these animals was drawn from the retro-orbital sinus under isoflurane anaesthesia as a terminal procedure. Blood samples were drawn into vacutainer tubes (BD) containing EDTA. Blood from two animals of the same genotype were drawn into the same tubes. Samples were gently mixed and then immediately centrifuged at 2000 g for 15 min at 4 C. Plasma was stored at À80 C until used.

Clinical cohort from Danish Center for sleep medicine
Approval for the use of blood samples was granted by the Danish Ethical Committee (KA03119). Thirty-seven subjects were included in this study. Lumbar punctures and blood sampling were performed for diagnostic purposes after informed consent from each subject. Lumbar punctures were performed between 8 and 12 AM and blood was drawn immediately after. Blood samples were drawn into Vacutainer tubes (BD) containing EDTA. Samples were gently mixed and then immediately centrifuged at 2000 g for 15 min at 4 C. Both CSF and plasma samples were stored at À80 C. Eighteen subjects diagnosed with NT1 were included in this study and 19 non-narcoleptic controls. Controls were selected from subjects seen at the sleep clinic that did not meet the criteria for any diagnosis. A summary of subject characteristics is listed in Table 1.
Clinical cohort from the Center for Narcolepsy of Bologna, Italy Approval for the use of blood samples was granted by the Comitato Etico Interaziendale Bologna-Imola (17009). The Italian cohort comprised 43 subjects, 21 subjects with an NT1 diagnosis and 22 controls. Unique to this cohort was that blood samples from narcoleptic patients were collected between 1 and 13 months from disease onset (8.4 6 4.2 months). Controls were selected from subjects seen at the sleep clinic that did not meet the criteria for a narcolepsy type 1 or type 2 diagnosis, 17 did not meet the criteria for any diagnosis while 5 had idiopathic hypersomnia. A summary of subject characteristics is listed in Table 1.

Assay development
Various combinations of temperature, heating time and buffer with varying pH were tested to investigate the effect of altering assay conditions on HCRT-1 signal from plasma and an HCRT-1 peptide standard. The pH of the buffers tested were pH 2 (phosphate buffer), pH 4 and 6 (acetic acid buffer), pH 8 (Tris:HCl buffer), pH 9 and 10 (sodium bicarbonate buffer). The concentration of all buffers used was 0.09 M. The heating temperature ranged from 40 to 90 C and heating time ranged from 0 to 120 min. SOP: 250 ml plasma was mixed with 250 ml Tris:HCl buffer and heated for 30 min at 65 C. HCRT-1 levels were measured by standard radioimmunoassay (RIA, #RK-003-30, Phoenix Pharmaceuticals).

Radioimmunoassay
RIA was used for quantification of HCRT-1 levels in all plasma, serum and CSF samples during assay development and testing of the clinical cohorts. All samples were run in technical duplicates. RIA kits were purchased from Phoenix Pharmaceuticals, CA, USA (#RK-003-30). The Phoenix antibody from the kit was used for detection unless a different antibody is specified. According to the manufacturer no cross-reactivity was observed with C-terminal HCRT-1 fragment (16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33), HCRT-2, Agouti-Related Protein (83-132)-amide, Neuropeptide Y, a-MSH or Leptin. Assay quality was monitored by the internal positive control sample included in the assay. Intra-assay variability was assessed by including a control sample from a pool of human CSF in each assay. The standard range of the RIA kit is 10-1280 pg/ml. RIA standard curves are depicted as B/B0% on the x-axis and concentration (pg/ml) on the y-axis. B0 ¼ total binding-nonspecific binding. B ¼ CPM of standard-non-specific binding. The plasma concentrations have been recalculated according to the initial plasma volume and are given in picograms per millilitres plasma. The person performing RIA was blinded to treatment conditions, group status and diagnosis.

High-performance liquid chromatography
High-performance liquid chromatography (HPLC) was used for chromatographic characterization of HCRT-1 in plasma/serum and HCRT-1 standard peptide. Samples were loaded onto an Xselect C18 column (Waters) and eluted with a linear gradient of acetonitrile containing TFA. Fractions were collected from Minute 2 to 20 (1 ml/min) and each fraction was dried by N 2 for 3 h and stored at À20 C until further use. Overview of the

Immunoprecipitation for HPLC analysis
Immunoprecipitation (IP) of plasma samples were performed according to the RIA protocol except that addition of 125 I-hcrt was omitted. In brief, SOP treated plasma was mixed with Phoenix antibody and incubated overnight at 4 C (22 h). After incubation Goat Anti-Rabbit IgG serum and Normal Rabbit Serum was added to the tubes. Tubes were vortexed and incubated at room temperature for 90 min. RIA buffer was added to the tubes and centrifuged at 1700 g for 20 min at 4 C. The supernatant was discarded and samples were stored at À20 C until use. Ten IP samples were dissolved in a total of 120 ml mobile phase A reagent (5% MeCN:95% H 2 O:0.1% TFA) and 99 ml of the solution was injected to the HPLC apparatus and run according to the HPLC program shown in Supplementary Table 2.

Testing of clinical cohorts
Prior to HCRT-1 quantification by RIA, plasma samples were pre-treated according to the developed SOP. RIA was initiated directly after pre-treatment. All plasma and CSF samples were analysed in technical duplicates. To ensure least amount of variation between samples, all samples were analysed in one RIA setup. The mean CV% was 2.5% and 2.2% for CSF and plasma, respectively. The mean CV% for the Italian cohort was 6.6% for plasma. The person performing pre-treatment of plasma, RIA and data analysis was blinded to diagnosis.

Data analysis
Results are reported as mean 6 SD. HCRT-1 plasma levels in NT1 and controls were compared using Student's t-test. For comparison of Hcrt-1 (mouse hypocretin-1) plasma levels of Hcrt-KO, Hcrt-Atxn3 and WT mice and during assay development a one-way ANOVA with a post hoc multiple comparison test was used. To access relationship between HCRT-1 levels in plasma and relevant variables of the clinical cohorts, we performed a multiple regression analysis. For the Danish cohort, variables considered for the analyses were HCRT-1 plasma levels, BMI, gender, age, diagnose, plasma storage time and albumin concentration in plasma. These variables were deemed appropriate based on previous literature. For the Italian cohort, variables considered for the analysis were HCRT-1 plasma levels, BMI, gender, age and diagnose.
For the sub-analysis of NT1 patients from the Italian cohort, HCRT-1 plasma and CSF levels, BMI, age, diagnose and time of blood sampling from disease onset were included. In each model, assumption of linearity, independence of errors, homoscedasticity and normality of residuals were tested. For statistical analysis, a P-value <0.05 was considered statistically significant (two-sided). All data analyses were done using the statistical program RStudio. 26

Data availability
Data are available upon reasonable request.

Results
Separation from protein carriers in blood aids in detection of HCRT-1 We hypothesized that the presence of HCRT-1 in blood is masked in the assays used today due to an interaction between HCRT-1 and protein carriers in blood. This is a common feature of steroid and lipophilic hormones transported in plasma. 27 Facilitating dissociation of HCRT-1 from such protein carriers should allow for detection of HCRT-1 in blood. To test this, we took advantage of the fact that the HCRT-1 standard peptide provided with the RIA assay used in the clinic, is mixed with albumin as protecting factor. Using this we performed western blot using antibodies detecting HCRT-1 and albumin on the HCRT-1/albumin mix and on CSF (Fig. 1A). Both samples were treated with two different buffers: (i) a 50 mM DTT-containing LDS reducing and denaturing sample buffer or (ii) a Leammli buffer with beta-mercaptoethanol. The blot shows strong immunoreactivity at the expected size of albumin, and HCRT-1 signal only in the sample treated with LDS buffer indicating that dissociation from albumin increases free HCRT-1 in the standard. No HCRT-1 signal was seen in the CSF sample demonstrating the low sensitivity of western blotting. RIA detection is well known to have a much higher sensitivity compared with western blot, and can detect HCRT-1 in CSF with a detection range of 40-500 pg/ml. The standard RIA protocol does, however, not show a reproducible HCRT-1 signal from crude plasma or serum. In most runs we did not see any signal from plasma or serum, but we occasionally did get a signal and in these cases the reproducibility was low. Instead, we speculated whether HPLC fractionation of plasma or serum to separate blood constituents based on physiochemical properties would allow for detection of HCRT-1. Indeed, HPLC fractionation of a serum sample followed by detection of HCRT-1 in the individual fractions by RIA revealed a positive signal in fraction 8 and 9 corresponding to what was obtained with the HCRT-1 standard (Fig. 1B). This shows that it is possible to detect HCRT-1 in blood samples following HPLC separation. Additionally, SDS-page confirmed that HPLC fractionation facilitated separation of HCRT-1 from albumin which eluted later than what HCRT-1 did (Supplementary Fig. 1A). Similar was observed from the HPLC chromatogram ( Supplementary Fig. 1B).
Taken together, our data suggest that dissociation of HCRT-1 from protein carriers allows for its detection in blood samples.
Drawbacks to HPLC fractionation were poor column lifetime and lack of reproducibility. After one HPLC run, the retaining properties of the column seemed lost. Plasma and serum samples indeed contain a significant amount of salt and proteins that can precipitate or adsorb on the packing material, eventually clogging the column. Incidence of clogging is high when crude plasma or serum is injected directly into the column as it was in these experiments. Simple sample pre-treatment, such as protein precipitation with acetic acid, prior to HPLC did not solve this problem since the HCRT-1 signal then was lost. The reason for this is likely that HCRT-1 precipitates together with its protein carriers. To avoid these problems, we explored other means of dissociating HCRT-1 from its protein carriers.

pH-change and heat treatment of plasma increases HCRT-1 signal
The strength of intermolecular interactions holding a ligandÀprotein complex together can be affected by a range of external factors such as pH, heat, hydrophobic surfaces, high shear or the presence of metal ions. Altering assay conditions can therefore significantly change the binding affinity of a ligandÀprotein complex and facilitate dissociation. We hypothesized, that a change in pH and/or heating of plasma would facilitate dissociation of HCRT-1 from protein carriers in the blood and that would in turn allow for detection of HCRT-1. We indeed found that changing the pH of the plasma solution (whilst heating the plasma sample) influenced HCRT-1 detection (Fig. 1C). At pH 6 no HCRT-1 signal was detectable, whereas a signal was observed in the pH range of 8-10 suggesting that basic pH provides the best conditions for HCRT-1 dissociation and detection. The highest HCRT-1 yield was obtained at pH 9 and 10, but due to increased background signal, we chose pH 8 for achieving the most reliable HCRT-1 signal. Similarly changing the pH of the HCRT-1 standard increased the HCRT-1 signal compared with untreated (À) HCRT-1 standard (Supplementary Fig. 3B). This is in agreement with the fact that the HCRT-1 standard contains albumin as carrier. Heating temperature also affected the HCRT-1 signal in plasma ( Fig. 2A). Below 60 C no HCRT-1 signal was detectable. The highest HCRT-1 yield was obtained with heating temperatures of 65 C and 70 C. Temperatures exceeding 70 C was accompanied by a decrease in HCRT-1 signal compared with the 65-70 C range, likely due to general protein precipitation or denaturation. Figure 2B shows the effect of heating time on HCRT-1 yield for plasma and HCRT-1 standard. Increasing heating time resulted in an increase in HCRT-1 yield. However, exceeding 30 min decreased the HCRT-1 signal from the HCRT-1 standard. Based on these initial experiments, a buffer of pH 8 and heating for 30 min at 65 C was chosen as the SOP. Figure 2C summarizes the effect of such treatment on the detectable signal from plasma HCRT-1 and HCRT-1 standard. Our data suggest that heat and not buffer is the main contributor to HCRT-1 dissociation and detection. No signal was observed if plasma was not exposed to heat.
CSF is a much less complicated matrix compared with blood, and in CSF HCRT-1 is readily measurable with RIA with no pre-treatment. This could be due to much lower protein binding of HCRT-1 in CSF as CSF albumin concentration is <0.5 g/l while it is 35-50 g/l in plasma. To investigate this, we tested the effect of SOP treatment on CSF HCRT-1 levels of four patients from the Danish clinical cohort. CSF samples from two NT1 patient with established CSF HCRT-1 level <10 pg/ml and two controls >160 pg/ml were tested. SOP treatment of the samples increased the HCRT-1 signal, but to a very low extend (Supplementary Fig. 4). Both untreated and treated CSF of NT1 patients were below the standard range of the assay, <10 pg/ml. This has two important implications. First the increased HCRT-1 signal observed following SOP treatment of plasma cannot be attributed to an increased antigenicity of HCRT-1 per se altering antibody binding. It also shows that our treatment does not cause a false positive signal from the formation of new non-HCRT-1 epitopes as a result of protein denaturation of proteins present in CSF.
Since the SOP treatment gave rise to an increased signal from the HCRT-1 standard with albumin as carrier, we wanted to test whether albumin itself, the most abundant protein in blood, would give rise to a false positive signal in our setup. We tested untreated and SOP treated BSA and recombinant human serum albumin. For all samples, no HCRT-1 signal was observed following treatment, suggesting that the increased signal following SOP treatment of the HCRT-1 plasma or standard is not caused by the presence of albumin (data not shown).
Previous efforts to measure HCRT-1 in plasma have mainly been based on solid phase extraction (SPE) of acidified plasma loaded onto a C18 columns prior to quantification using either RIA or enzyme immunoassays. However, in our hands this approach showed varying results and low reproducibility. Figure 2D shows two SPE experiments and a SOP treatment experiment with the same plasma pool. Our assay gives rise to a higher and a more reproducible signal compared with the conventional SPE protocol. Increasing plasma volume to 1000 ml in the SPE protocol did not give rise to a higher signal than 250 ml further questioning this method due to lack of linearity.
Assay linearity in our case was tested in a linearity-ofdilution experiment, where seven different plasma dilutions were included (Supplementary Fig. 5A). A linear relationship between the observed HCRT-1 levels in plasma and the expected HCRT-1 levels was established with a correlation coefficient of 0.991, indicating that the assay has little to no matrix interference.
Assay interference by plasma anticoagulants and serum were finally evaluated to select the best blood collecting system for blood sampling as was sensitivity to freezing. Citrate-, EDTA-and heparin plasma plus serum was collected at the same time from two healthy individuals, and half of each sample was frozen for 30 min at À80 C. The samples were next SOP treated and HCRT-1 levels were quantified (Supplementary Fig. 5B). None of the anticoagulants nor serum significantly changed the detected HCRT-1 levels indicating that assay interference by the type of blood collection system is low for this assay.

Validation of specificity of HCRT-1 immunodetection after sample pre-treatment
To confirm the presence of genuine HCRT-1 in SOP treated plasma samples, we performed IP of SOP treated plasma with the Phoenix antibody from the RIA kit and performed HPLC separation of the sample followed by detection of HCRT-1 in the HPLC fractions (Fig. 3A). As expected from previous HPLC runs (Fig. 1B), HCRT-1 was detectable in fraction 8 (Fig. 3A). It was necessary to pool 10 IP samples to get a detectable signal confirming the very low levels of HCRT-1 in plasma. Figure 3A also shows the HCRT-1 concentration following SOP treatment or no treatment of the same plasma sample used for IP/HPLC.
To further validate the HCRT-1 RIA signal, we tested three different commercially available HCRT-1 antibodies ability to detect plasma HCRT-1 in RIA settings. All three antibodies were able to detect HCRT-1 in plasma following SOP treatment but not in crude plasma (Fig. 3B). However, the concentration of HCRT-1 in SOP treated plasma varied depending on the antibody used. Interestingly, when measuring HCRT-1 in CSF of a healthy control and Narcoleptic Type 1 patient using the three antibodies, markedly different levels of HCRT-1 were found for the control subject ( Supplementary Fig.  5A). To further explore the specificity of the antibodies, the Phoenix HCRT-1 peptide in the RIA setting was replaced with a modified HCRT-1 peptide with an extra glutamic acid residue in second position to disrupt N-terminal binding (Supplementary Fig. 6B). A clear difference was observed between standard curves manufactured with the two different peptides ( Supplementary  Fig. 6C and D). The Phoenix and Pierce antibodies both showed a much lower binding affinity to the modified peptide. The Peninsula antibody was however still able to bind the modified peptide with high affinity. These data confirm that the Phoenix and Pierce antibodies bind to the N-terminal region of HCRT-1, while the Peninsula antibody recognize a different epitope.
We also measured Hcrt-1 in mouse blood using the two other HCRT-1 antibodies mentioned above. With the Pierce antibody, we also saw a significantly decreased signal from Hcrt-KO mouse blood ( Supplementary Fig.  7), while the Peninsula antibody gave similar signals from WT and Hcrt-KO mouse blood. This confirms what we report above, that while the Phoenix and Pierce antibodies show specificity towards the same HCRT-1 epitope, the Peninsula binds to something else in addition. This shows that the Peninsula antibody is not a good antibody for measuring HCRT-1 levels. The Pierce antibody seems to be highly specific and very useful for HCRT-1 detection. Unfortunately, this antibody was discontinued by the vendor limiting the number of experiments we were able to perform with this antibody.

HCRT-1 levels are similar in blood samples from narcolepsy type 1 patients and controls
To explore the feasibility of our assay to be used as a diagnostic tool for identifying NT1 patients, we measured HCRT-1 plasma levels in samples from two independent clinical cohorts. Plasma levels of HCRT-1 in 18 NT1 (Danish cohort) with confirmed low HCRT-1 levels in CSF (Fig. 4A) were similar to the levels detected in a matched control group (NT1: 244.2 6 17.25 pg/ml plasma and controls: 240.9 6 28.78 pg/ml plasma, ns) (Fig. 4B). Minimum and maximum values in the cohort were 195.2 and 310.7 pg/ml plasma. To elucidate possible factors affecting plasma HCRT-1 levels, we performed a multiple regression analysis with age, gender, body mass index, diagnosis, albumin concentration and storage time as predictors ( Table 2). None of the predictor variables showed a significant P-value, suggesting that the strength of the effect on plasma HCRT-1 concentration is low for all variables. Our finding of normal HCRT-1 levels in NT1 patient plasma was replicated in an independent clinical cohort from Bologna, Italy. Again HCRT-1 levels in plasma did not differ between the two groups (NT1: 209.8 6 23.3 pg/ml plasma and controls: 209.2 6 23.65 pg/ml plasma, ns) (Fig. 4D). Minimum and maximum values in the Italian cohort were 165.8 and 244.8 pg/ml plasma. For the multiple regression analysis, age, gender, body mass index and diagnosis were tested as predictors. None of the predictors had an effect on HCRT-1 plasma levels in this model (Table 2).
We also measured HCRT-1 levels in NT1 patients and controls in the Italian cohort by use of a different HCRT-1 antibody (the Pierce antibody) and confirmed our finding of similar HCRT-1 levels between the two groups ( Supplementary Fig. 7). In contrast to the Danish cohort, the Italian NT1 blood samples were obtained close to onset of disease. To evaluate the effect of disease length on HCRT-1 plasma levels in these patients, we performed a multiple linear regression analysis where time of blood sampling after disease onset (months) was included as predictor. The analysis showed a significant effect of age and time of blood sampling after disease onset. HCRT-1 increased with time after disease onset but at the same time showed a small decrease with age. The overall F-test showed that this model, at a significance level of 0.05, can predict plasma HCRT-1 level [F(4,16) ¼ 3.18, P ¼ 0.042]. When HCRT-1 levels were measured with the Pierce antibody, there were no significant effects of neither age nor time of blood sampling (Supplementary Tables 1 and 2).

Discussion
In this study, we have developed a simple reproducible assay to evaluate HCRT-1 levels in blood. We found that heating plasma at pH 8 enabled reproducible detection of HCRT-1. Using HPLC, different antibodies and mice lacking hypocretin, we confirmed the HCRT-1 signal from blood samples, but we also showed that the signal: background ratio in our assay is low-in the order of 1:7. From our measurements on blood samples from mice lacking hypocretin we estimated that HCRT-1 is present in plasma in a concentration of $32 pg/ml. Using our assay, we could measure identical signals of HCRT-1 immunoreactivity in plasma samples from NT1 patients and controls in two independent clinical cohorts. Additionally, no correlation was observed between blood and CSF HCRT-1 levels in any of the two groups. The data presented here suggest that peripheral HCRT-1 is intact in NT1 and thus blood HCRT-1 measurements cannot be used as diagnostic tool to identify NT1. It further suggests that the NT1 disease process primarily involve CNS HCRT neurons whereas potential peripheral HCRT producing cells are intact.
We suggest that the increase in signal we see following heating of plasma to 65 C is due to release of HCRT-1 from its blood carriers. Albumin, the most abundant carrier protein in the blood, has indeed been shown to go through a phase shift at this temperature. [28][29][30] In addition, it has been shown that a vast majority of the low molecular weight biomarkers detectable in blood are bound to large carrier proteins, representing a subproteome itself. Mass spectrometry analysis has shown that removing high molecular weight proteins also removes a substantial amount of the low-molecular weight proteins. 31,32 This could explain why so far, no mass spectrometry-based method has been developed for measuring HCRT-1 in blood, as it is a common procedure to remove proteins from plasma, before attempting peptide detection with mass spectrometry.
A limitation of our assay is the high unspecific background signal. This is likely caused by unspecific binding of the antibodies to blood proteins that change antigenicity from denaturation during the heating step. The unspecific signal does not come from albumin, as heating albumin alone in isolated form did not produce a signal. Heating of CSF did produce a slight increase in the signal from our assay, but not at all to the extend we see in blood. We did try different filtering steps to bring this background signal down, but with any filtering the HCRT-1 signal was also lost. This is likely because HCRT-1 is very hydrophobic and sticks to the filter/tubes etc.
We expect that the background noise is in the same order in mice and humans. The estimation of blood Hcrt-1 levels around 30 pg/ml would thus apply to both mice and humans. Since the normal HCRT-1 RIA has a detection level for reliable detection at 40 pg/ml, this explains why HCRT-1 cannot be detected in blood by RIA without pre-treatment.
Our finding of intact peripheral HCRT-1 in NT1 raises numerous questions regarding the origin of peripheral HCRT-1, as well as the aetiology of NT1. Importantly, the data suggest that peripheral HCRT-1 does not solely originate from the CNS and that there instead is a peripheral source of HCRT-1. Intact peripheral HCRT-1 in NT1 would mean that peripheral HCRT-1 producing cells are left unaffected by the autoimmune attack that destroys the HCRT neurons in the brain, the hallmark of NT1 pathology. The implication of this would be that HCRT itself is not the target of the autoimmune attack. Another protein expressed in the HCRT neurons but not the peripheral producers would in this case be the primary target, resulting in a selective destruction of the central HCRT producing neurons. A study of cytotoxic T cells targeting proteins in HCRT neurons did indeed suggest other factors than HCRT-1 to be the autoimmune target, 33 but other papers have demonstrated direct targeting of HCRT-1. [34][35][36] So far, no studies have demonstrated direct in vivo targeting of HCRT neurons, so the nature of the primary autoimmune target in NT1 remains unknown. Multiple linear regression analyses showed no significant effect of any of the predictors in the two cohorts. However, sub-analysis of the NT1 patients in the Italian cohort revealed a significant effect of the time of blood sampling after disease onset and age. This model could significantly predict plasma HCRT-1 level. Level of significance at <0.05 was considered statistically significant.
An alternative explanation to our finding could be that the HCRT-1 producing neurons in hypothalamus are more sensitive and prone to death following the attack, whereas peripheral HCRT-1 producing cells might recover more easily. This could result in low blood HCRT-1 levels at disease onset and then a recovery over time. A low HCRT-1 blood level could in this case be indicative of an active autoimmune process. We did see a correlation between time from disease onset and plasma HCRT-1 levels in our Italian cohort with lower levels closer to onset. This effect was, however, small, and more studies are needed to determine if this is a reproducible effect.
No significant correlations were observed between HCRT-1 plasma concentration and BMI, age, gender, albumin concentration, diagnosis or CSF HCRT-1 concentration in any of the cohorts. These findings contrast with previous studies showing a correlation with a number of these factors (Supplementary Table 1). However, discrepancies among these previous studies are found as well. Adam et al. 37 report lower blood HCRT-1 levels in obese compared with lean subjects. Baranowska et al. 38 and Tomasik et al. 39 report the same findings in women and children during puberty, respectively. In contrast, Heinonen et al. 40 observed significantly higher HCRT-1 levels in obese subjects. These discrepancies may be explained by the different assays used, the sensitivity of the assays and potentially the use of different antibodies. In the present study, we observed significantly different HCRT-1 levels in both blood and CSF when measuring the same sample with different antibodies. Using Phoenix and Pierce antibody we found that the HCRT-1 levels in blood of the controls were lower than CSF HCRT-1 levels, which is in contrast to what we observed with the Peninsula antibody. With the Peninsula antibody we observed surprisingly high HCRT-1 blood levels, while measuring CSF HCRT-1 levels with the Peninsula antibody showed lower CSF HCRT-1 levels in comparison. We speculate that these differences might arise from the Peninsula antibody having a strong cross-reactivity with other components of plasma. The difference in signal could also be due to modified or degraded variants of HCRT-1 in blood or CSF. It has previously been shown that fractions of HCRT-1 are present in CSF, 41 but whether this is true for plasma is not known. Importantly, our data demonstrate that one should take caution when comparing HCRT-1 levels detected with different antibodies/assays.
Despite the lack of a relationship between disease state and HCRT-1 plasma levels in NT1, our assay provides a solid platform for future studies on the peripheral actions of HCRT-1. Measuring HCRT-1 levels in blood could be relevant in a range of diseases besides NT1 and might help determine the role of peripheral HCRT-1. The narrow concentration range of HCRT-1 in blood could suggest a tight regulation, supporting the idea of HCRT-1 as a peptide hormone. Where peripheral HCRT is produced and how this is regulated is still unknown and needs further attention.

Supplementary material
Supplementary material is available at Brain Communications online.

Funding
The project was funded by an investigator driven grant from UCB Pharma and by Proof of Concept (PoC) funding from Region Hovedstadens Tech Transfer Office.