Introduction

Sweat gland activity plays an important role in thermoregulation in humans by enabling through its secretory activity evaporative water loss in the skin surface1. In great part the efficiency of this system is contingent on tight regulation of electrolyte composition to avoid excessive losses that could be potentially life threatening. The best example of this is the disease Cystic Fibrosis (CF), serendipitously recognized during a heat wave as being characterized by an abnormality in sweat electrolyte composition2. This aspect of sweat gland physiology has been the focus of intense investigation, establishing the sweat chloride concentration determination as the gold standard test for the diagnosis of CF3. Thanks to the seminal work of Quinton, among others, an impermeability to chloride transport introduced by the defective CFTR gene function characteristic of CF was recognized as the basic defect in this disease4.

Sweat is recognized to also contain a large number of additional solutes, including proteins, lipids, metabolites and other small molecules5. This rich composition suggest the sweat gland to also play an important role in other aspects of skin surface homeostasis, including barrier integrity, antibacterial properties and moisture preservation6. This also suggest that the concentration in sweat of a given analyte could also reflect a specific condition in an individual. Hydration status is a good example of a physiologic state that could be monitored through sweat composition, given its recognized electrolyte composition regulation7. Thus, sweat has been proposed as an alternative biofluid that could be used instead of blood sampling to use it as a proxy for the circulating concentrations of metabolites, biomarkers and drugs of interest. For this reason there is a great interest in exploring the use of sweat measurements as a reflection of health status.

Although sweat seems appealing as an accessible biofluid for non-invasive assessment, it carries a number of challenges to put into practice8. For the most part, obtaining a sweat sample will entail the stimulation of secretion by the glands to access a large enough volume, the collection of the sample and the measurements of the specific analyte of interest. Multiple technological advances at hand are now allowing for highly efficient performance on all this aspects, and even enabling repeated and/or continuous sampling and monitoring. However, perhaps a more important aspect that is frequently ignored is the physiology behind the appearance of a given analyte in sweat samples. This includes the simple question of the actual source, since sweat is typically collected from the skin surface and simple losses through the stratum corneum could play a role.

Among analytes of interest are free aminoacids and their metabolites. Aminoacids are the building blocks of proteins and their concentrations in circulation are frequently used to carry metabolic profiling9,10,11,12,13,14,15,16. In addition, aminoacids have been proposed as biomarkers for a variety of disorders17,18,19,20,21,22,23,24,25. This also includes the diagnosis of inborn metabolic disorders, as well as the monitoring of the response to therapy and dietary intervention for these disorders26. There are multiple examples in the literature of technological developments for the sampling of minute amounts of sweat, sample preparation and detection of aminoacid concentrations in sweat27,28,29,30,31,32,33,34,35,36,37,38,39. However, few reports on the concentrations of aminoacids in sweat of healthy individuals point to a high variability as well as variable correlations with circulating levels40. We wanted to further explore the potential use of aminoacids in sweat for health monitoring by evaluating the concentrations in relation to the concurrent concentrations in circulation. We also wanted to evaluate for the potential confounding effect of non-sweat gland sources by carrying collections onto filters to eliminate cellular debri as a potential contaminant source, as well as applying an impermeable oil barrier to collect pure sweat gland secretion.

Methods

Human subjects

We recruited healthy adult volunteers from a pool of eligible subjects at Stanford University. Eligibility criteria included male or female older than 18 years, in good health, with no history of dermatologic disease or dietary restrictions, free from acute illness within 2 weeks of study participation, no allergy to cholinergic agents, and not taking medications with anticholinergic properties. Subjects that passed screening for study entry criteria were instructed to refrain from the use of any lotions or skin care products and were scheduled for a study visit at least week later. All research activities were performed in accordance with relevant institutional guidelines and regulations. The study protocol was approved for human subject participation by Stanford University’s Human Subjects Institutional Review Board (IRB protocol#59109). All participants provided written informed consent prior to any study procedures. This included provisions for sample collection procedures, and banking of specimens and data.

Sweat sampling

Sweat samples were collected following a well-established standard operating procedure that includes steps for sweat gland secretion stimulation, collection of sweat, and retrieval of sample. Samples were handled following specific precautions to avoid sample loss from incomplete retrieval or evaporation. In brief, the volar surface of the forearm was wiped clean with alcohol swabs and rinsed twice with deionized water prior to stimulation by pilocarpine iontophoresis. This was performed with a Wescor® iontophoresis device with pilogels (ELITech group Inc., Logan, UT) and following manufacturer instructions. This was followed by rinsing twice the stimulated area with deionized water and then drying the skin surface with air. We then placed 10 hygroscopic filter discs (0.5 inch diameter, Whatman 10400212) to cover the stimulated surface completely. A square of Tegaderm® patch was then applied to seal the stimulated area with the filters. After 30 min, the Tegaderm® was carefully peeled back on a corner and the filters were quickly removed and transferred to a centrifugal filter tube (Ultrafree MC SV centrifugal unit, Millipore®). In a last step, the sweat sample was extracted from the filter discs by centrifuging the filter tube at 13,000 g for 10 min at 4oC. The collected sweat sample was transferred to a cooled PCR tube, tightly sealed and stored at -80oC for later batch analysis.

Blood sampling

Prior to sweat collection a 10 mL blood sample was obtained by venipuncture of a peripheral arm vein into EDTA tubes. The sample was then centrifuged at 600 rpm to separate the plasma. Plasma samples were transferred to 0.5 mL cryovials, tightly closed and stored at -80oC for later batch analysis.

Aminoacid analysis

Concentrations of 22 free aminoacids were assayed at Metabolon Inc. (Metabolon Inc. Morrisville, NC) following a standardized protocol. In brief, samples were spiked with stable labelled internal standards and subjected to protein precipitation by treatment with an organic solvent. After centrifugation, an aliquot of the supernatant was diluted for injection into the measuring instrument. Measurements were performed by LC-MS/MS on an Agilent 1290 UHPLC/Sciex QTrap 5500 system equipped with a C18 reversed phase UHPLC column. The mass spectrometer is operated in positive mode using electrospray ionization (ESI). The peak areas of the individual analyte parent ions were measured against the peak areas of the parent ions of the corresponding internal standards. Quantitation was performed using a weighted least squares regression analysis generated from fortified calibration standards prepared immediately prior to each run. We also recorded specific properties of each aminoacid to evaluate the role they could play in their partitioning into sweat. This included their molecular, with, polarity, charge and LogP. The logP of a molecule represents its hydrophobicity, and we have previously shown this to be a characteristic of importance for certain drugs41.

Statistical analysis

Data are presented as means and SD. Comparisons between blood and sweat concentrations were performed individually for each aminoacid of interest by paired t-Test. Correlations between blood and sweat were performed by Pearson correlation, and for those found significant, predictive ellipse plots with 90 and 95% confidence limits were generated. For all comparisons the level of significance was a set a priori at an α of 0.05. All statistical analyses were performed with the SAS statistical package (v.9.4, SAS Institute, Cary NC).

Results

A total of 48 adult subjects (36 females and 12 male) with a mean (SD) age of 38.3 (12.8) years were enrolled and completed sampling procedures. Of these, 7 sweat samples were measured at < 100 µL in volume and considered insufficient for further testing. Thus, complete analysis was performed in samples from 41 subjects. Most samples had detectable aminoacid levels above the lower limit of detection determined for this assay for each aminoacid (range 0.05 to 2.5 µg/mL). Only 2 subjects had one aminoacid not detected in their plasma sample (Asparagine), but both subjects had this aminoacid detected in sweat. For sweat, only 4 subjects had one aminoacid not detected: 2 for trans-hydroxyproline, 1 for arginine and 1 for methionine, but these were detected in their plasma samples.

Figure 1 presents Box plots of the distribution of concentrations of the aminoacids measured in plasma and sweat. Overall, only for Glutamate, Ornithine and Valine weak correlations were noted between their plasma and sweat concentrations (Pearson Rho coefficients ~ 0.3 and p < 0.1). Interestingly for all 3 this was a negative correlation, which will suggest less transfer of the aminoacid into sweat as the plasma concentration increases. Table 1 presents the mean values measured for each aminoacid, as well as their molecular weight and LogP value. We found the mean concentrations of Leucine, Phenylalanine, Proline, Tryptophan and Valine to be comparable between Plasma and Sweat (all p > 0.2). Only for glutamine, Lysine, Methionine and Trans-Hydroxyproline we find the concentrations being significantly higher in plasma than sweat (all p < 0.0001). For the remainder 13 the concentrations were significantly higher in sweat than plasma (all p < 0.0003).

Fig. 1
figure 1

Box plots of the distribution of aminoacids in plasma (blue) and sweat (red) concurrently measured in healthy volunteers (n = 41).

Full size image
Table 1 Molar concentrations (mean ± SD) for aminoacids measured in concurrent samples of plasma and sweat, along with their molecular weight (MW) and logp, with p-values for comparisons between the means by paired t-test.
Full size table

To clarify the potential role that transdermal loss instead of actual gland secretion may play on the partitioning of aminoacids into the skin surface, we collected concurrent samples of plasma and sweat on a subset of 24 healthy volunteers from our original cohort. For this set of sweat sample collections, we stimulated sweat gland secretion by pilocarpine iontophoresis as detailed above, but the sweat collection was performed under two different conditions in the same subject. On one forearm it was collected as described above, and on the contralateral arm a thin layer of mineral oil was applied to the stimulated area prior to placing the filter discs. By providing an impermeable barrier, we attempted to minimize the transdermal loss and allow only secreted sweat to be collected onto the filters as we have shown before42. Then each subject provided one plasma sample and two sweat samples under the conditions described (under a barrier and without a barrier). Analysis of these samples for the 25 aminoacids previously assayed, found only 15 detectable above the lower limit of quantification in all collected plasma and sweat samples, whether with or without an oil barrier. Then, aminoacids detected for this analysis included: Alanine, Arginine, Asparagine, Glutamine, Glycine, Isoleucine, Leucine, Methionine, Phenylalanine, Proline, Serine, Threonine, Tryptophan, Tyrosine, and Valine. Overall, in this second set of samples we noted for most aminoacids similar concentrations, though with larger variability. As detailed in Table 2, only for Alanine, Asparagine, Glycine, Serine, and Threonine we noted a significantly lower concentration in the sweat collected under a barrier. This will suggest that for these aminoacids there is an important transdermal source for appearance into the skin surface.

Table 2 Molar concentrations (mean ± SD) for aminoacids measured in a second cohort of healthy volunteers in concurrent samples of sweat collected with and without a skin mineral oil barrier, with p-values for comparisons between the means by paired t-test.
Full size table

When we evaluated for correlations of the aminoacid concentrations between the concurrently collected plasma and sweat collected under an oil barrier, representing gland true secretion, we found a significant negative correlation (Fig. 2, Pearson Rho coefficient 0.49, p = 0.015) only for Valine, and weak correlations for Glycine, Leucine, Tryptophan and Tyrosine (Pearson Rho coefficients range 0.35 to 0.38, and p values 0.05–0.10). Interestingly, we found that for Alanine and Threonine significant positive correlations between plasma concentrations and the difference in concentrations between sweat collected without and with a barrier (Fig. 3, Pearson Rho coefficients of 0.63 for both, p < 0.0010).

Fig. 2
figure 2

Scatter plot of the distribution of Valine concentration in sweat collected under an oil barrier and concurrent plasma. There is a significant negative correlation with Pearson Rho coefficient of 0.489 (p = 0.015). Ellipses represent the predictive 90% and 95% confidence limits.

Full size image
Fig. 3
figure 3

Scatter plot of the distribution of Alanine (left panel) and Threonine (Right panel) concentrations between differences in the estimated transdermal source and concurrent plasma. There are significant positive correlations with Pearson Rho coefficient of 0.63 for both (p = 0.0010 fro Alanine and p = 0.0009 for Threonine). Ellipses represent the predictive 90% and 95% confidence limits.

Full size image

Discussion

Sweat collection has been a long standing procedure used to analyze chloride concentrations for the diagnosis of cystic fibrosis3. Sweat also contains metabolites such as amino acids which potentially could be used as a non- invasive biomarker to assess disease status, biometrics for individual identification, phenotypically profiling individuals with specific metabolic disorders, as well as providing the means for disease monitoring while providing insights into their individual health14,43,44,45,46,47. In addition, aminoacid profiles can provide valuable information on the metabolome associated with specific disorders48. As an example, we previously reported on important differences in aminoacid flux between healthy controls and subjects with cystic fibrosis reflecting an abnormal catabolic state49.

By using the standard sweat collection procedure, commercially available stimulation and collection devices, we were able to identify plasma and sweat amino acid correlations in a cohort of healthy controls. Overall, in the initial cohort analysis, Glutamate, Ornithine and Valine demonstrated a weak negative correlation between plasma and sweat. The correlations are so weak that it is difficult to speculate on transport mechanisms which for sweat range include passive diffusion, passive diffusion followed by ionization or deionization, or active transport either into or out of sweat50. Mean concentrations of Leucine, Phenylalanine, Proline, Tryptophan and Valine were comparable between plasma and sweat. For Glutamine, Lysine, Methionine and Trans-Hydroxyproline, the concentrations were consistently higher in plasma than sweat. Again, the correlations are too weak to suggest what modes of transport exist into or out of sweat. Further, unlike other studies for drugs in sweat41, we did not find characteristics such as polarity or LogP as differentiating those aminoacids with increased appearance in sweat. An aspect that seems of greater importance is simply their molecular weight. For both the plain sweat vs. plasma and sweat under oil vs. plasma there are significant correlations (Fig. 4), reflecting the smaller molecules having an easier time partitioning into sweat.

Fig. 4
figure 4

Plot of the individual aminoacid concentration difference between sweat collected without (right panel) or with an oil barrier (left panel) and concurrent plasma concentration. A significant negative correlation is noted, with larger differences for the smaller molecules (r = 0.74 and 0.76 respectively, both p = 0.002).

Full size image

An additional possibility to consider is that the induction procedure itself may have stimulated active secretion which will favor aminoacids with certain specific characteristics. We noted no clear pattern being apparent based on aminoacid polarity or logP, which will argue against this possibility. It also possible that the inotophoresis procedure itself may have induced some degree of injury altering sweat gland permeability or other factors caused by local inflammation after stimulation. Besides some minor brief redness in the stimulated area, which is likely result of transient vasodilation, there was no evidence of skin irritation or leakiness in the stimulated area in any of the subjects. The differences in the collections comparing the use of a barrier we interpret as a reflection of a transdermal source contaminating the sweat sample, which will then represent a significant component of the concentrations of these aminoacids in the collected sweat42. It is known that aminoacids are an important component of the humectants of the stratum corneum, with their source being product of the degradation of epidermal S-100 proteins51,52,53. Then the differences noted will represent the amount transferred into sweat from transdermal loss. The data reveals to us that aminoacids are more complex in their partitioning into sweat than previously assumed. By extension, this suggests to us that other endogenous small molecules may have limited diagnostic value in sweat compared to other biofluids like blood or ISF. Amino acids are small molecules, so researcher ambitions to measure larger peptide and proteins should therefore be treated with even greater caution given larger molecules in theory should have greater difficulty partitioning into sweat and therefore even worse blood correlation54.

Our study has a few limitations. First, we did not monitor directly the secretory rate induced in the stimulated area. It is unclear what effect the secretory rate could have on aminoacids, but work on electrolyte concentrations suggest that secretory rates can have a significant effect. Previous study showed aminoacid profiles comparable to those obtained after exercise55. However, the stimulation with pilocarpine at the doses utilized is known to maximally stimulate glands, so our results reflect the concentrations under maximal secretory activity and then it is unlikely secretory rate was a factor in our results. Future studies will evaluate the effect of lower and varying secretory rates, which could be accomplished by manipulating the stimulation conditions. Second, our results possibly do not reflect real world situations since we induced sweating by stimulation with pilocarpine. It is possible that concentrations could differ, as studies done by collections under exercise conditions or thermal stimulation suggest high variability even within subjects. This will also have to be taken into consideration for future studies, since it is possible that the metabolic conditions imposed by exercise could by themselves also have an effect on aminoacid flux. Third, for some polar aminoacids contamination of the plasma sample with the chelating activity of EDTA could have interfered with the measurement assay. However, for the most part we feel this possibility of a potential interference was low and likely mitigated by collecting an adequate volume of blood into the tubes used and per the fill volume recommended by the manufacturer. This likely minimized any potential carryover of EDTA into the plasma that could have interfered with the assays performed56. Lastly, to what degree dietary load could influence aminoacid metabolism was also not contemplated. This in particular when it comes to the intake of essential aminoacids, which under low intake or deficiency conditions could have an effect on overall aminoacid flux44.

If amino acids in sweat are to be used as a biomarker, we need a better understanding as to the degree to which their concentration is a true representation of gland secretion and the mechanisms behind their secretory responses. Perhaps this may call for additional in vitro work with intact glands, as was done for electrolytes1,4. The differences noted in concentrations and correlations between plasma and sweat collected with and without the oil barrier suggest to us that collection conditions are also an important aspect that is often ignored42. This will be important to bring into consideration as innovative platforms for sampling continue to be advanced.We find it of interest that when we evaluated for correlations of concentrations between the concurrently collected plasma and sweat under an oil barrier, representing gland true secretion, we found only for Valine a significant correlation. In addition, for Alanine and Threonine the difference introduced by the transdermal source was correlated with plasma suggesting that their correlation with plasma could be modeled taking into account the collection conditions or controlled by collection device design. This suggests to us that these amino acids might be more amenable to development as surrogates of circulating levels. Further investigation in additional healthy controls and disease cohorts may provide insight into the utility of monitoring these aminoacids.