ORIGINAL RESEARCH published: 25 May 2020 doi: 10.3389/fphys.2020.00373 Proteomic Profiling and Monitoring of Training Distress and Illness in University Swimmers During a 25-Week Competitive Season Amy M. Knab 1* , David C. Nieman 2 , Laura M. Zingaretti 3 , Arnoud J. Groen 4 and Artyom Pugachev 4 1 Department of Kinesiology, Queens University of Charlotte, Charlotte, NC, United States, 2 North Carolina Research Campus, Appalachian State University, Kannapolis, NC, United States, 3 Centre for Research in Agricultural Genomics, Barcelona, Spain, 4 ProteiQ Biosciences GmbH, Potsdam, Germany Purpose: To evaluate relationships of proteomics data, athlete-reported illness, athlete training distress (TDS), and coaches’ ratings of distress and performance over the course of the competitive season. Methods: Thirty-five NCAA Division II swimmers were recruited to the study (male Edited by: Beat Knechtle, n = 19, female n = 16; age 19.1 ± 1.6 years). Athletes provided fingerprick dried blood University Hospital Zürich, Switzerland spot (DBS) samples, illness symptoms, and TDS every Monday for 19 of 25 weeks in Reviewed by: their season. Coaches monitored performance and rated visual signs of distress. DBS Richard J. Simpson, samples were analyzed for a targeted panel of 12 immune-related proteins using liquid The University of Arizona, United States chromatography/mass spectrometry (LC/MS). Henrique Pereira Neiva, University of Beira Interior, Portugal Results: Thirty-two swimmers completed the protocol. The data were grouped in 2– Argyris G. Toubekis, 3 weeks segments to facilitate interpretation and analysis of the data. TDS scores National and Kapodistrian University varied between athletes, and were highest during the early fall conditioning ramp up of Athens, Greece Jason Vescovi, period (8.9 ± 1.6 at baseline to a peak of 22.6 ± 2.0). The percent of athletes University of Toronto, Canada reporting illness was high throughout the season (50–78%). Analysis of TDS using *Correspondence: Principle Component Analysis (PCA) revealed that 40.5% of the variance (PC1) could Amy M. Knab knaba@queens.edu be attributed to illness prevalence, and TDS scores for the athletes reporting illness and no illness were different across the season (P < 0.001). The coaches’ ratings of Specialty section: swim performance and swimmer’s distress, sex, and racing distance (sprinters, middle This article was submitted to Exercise Physiology, distance, long distance) were not correlated with PC1. Linear Discriminant Analysis a section of the journal (LDA) analysis of the data showed a separation of the baseline weeks from exam weeks Frontiers in Physiology with or without competitions, and with competitions alone (p < 0.001). Seven of the 12 Received: 10 June 2019 proteins monitored over the course of training were upregulated, and the addition of the Accepted: 30 March 2020 Published: 25 May 2020 protein data to LDA analysis enhanced the separation between these groups of weeks. Citation: Conclusion: TDS and illness were related in this group of 32 collegiate swimmers Knab AM, Nieman DC, Zingaretti LM, Groen AJ and throughout the competitive season, and expression of immune proteins improved the Pugachev A (2020) Proteomic statistical separation of baseline weeks from the most stressful weeks. TDS data Profiling and Monitoring of Training provided by the swimmers did not match their coaches’ ratings of distress and swim Distress and Illness in University Swimmers During a 25-Week performance. The importance of the immune system in the reaction to internal and Competitive Season. external stress in athletes should be an area of further research. Front. Physiol. 11:373. doi: 10.3389/fphys.2020.00373 Keywords: swimming, proteins, inflammation, upper respiratory tract infection, mental stress Frontiers in Physiology | www.frontiersin.org 1 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers INTRODUCTION Current advancements in the field of proteomics may provide a mechanism to identify target biomarkers for overreaching in The linkage between intensive exercise training and increased athletes. Proteomics involves the large scale measurement of risk for illness has been an active area of research during the structure and function of proteins in a tissue or organism, the past several decades. Studies indicate that illness risk may and is useful in the identification of candidate biomarkers be elevated when an athlete engages in recurring cycles of for various disease processes and drug treatments (Rodríguez- unusually intense exercise training coupled with competitive Suárez and Whetton, 2013). Even though proteins are the main events (Nieman, 2000; Nieman and Wentz, 2019). In addition, components of the metabolic pathways of cells, proteomics, until periods of overreaching for any athlete, although necessary recently, has seldom been used in exercise-based, human studies for further adaptations in performance, come with significant (Balfoussia et al., 2014). In a prior study from our research group, physiologic changes that may compromise the immune system. endurance athletes served as their own controls and in random, Swimmers are an interesting cohort of athletes that face many counterbalanced order either exercised intensely for 2.5 h, 3 days barriers to performance including balancing training with in a row, or sat in the lab (Nieman et al., 2018). Fingerprick outside stressors. samples for dried blood spot samples (DBS) were collected from High-level university swimmers for example, participate in study participants before and after laboratory-based exercise or 10–15 competitions during the fall and winter seasons when rest sessions, and then during two recovery days. From this pilot pathogen exposure is high, train 3–5 h daily, experience heavy study, thirteen proteins were linked to functional overreaching travel schedules that disrupt sleep and eating habits, and undergo (FOR), and most were associated with the acute phase response high mental stress from educational and personal commitments and innate immune system activation. (Knab et al., 2013; Hellard et al., 2015; Collette et al., 2018). As a Thus, the purpose of this study was to use the recently result, illness risk can be elevated, interfering with the capacity to developed targeted panel of proteins in a group of high- train and race (Gleeson et al., 1995, 2000; Fricker et al., 2005). level swimmers over the course of the entire season to track A 4-year study of 28 elite swimmers showed that the odds for physiologic changes that may provide biologic markers of upper respiratory tract infection (URTI) increased 1.08 and 1.10 internal and external stress. In order to create a full pictures of times for every 10% increase in resistance and high-load training, these changes over the course of the competitive swim season, respectively (Hellard et al., 2015). various other measures were collected and included in the In response, several athletic organizations including the statistical model: training workload and performance assessment, International Olympic Committee (IOC) initiated acute illness quantified feedback from coaches, and perceived training distress surveillance systems to delineate the extent of the problem and and illness. We hypothesized that the use of blood sample underlying risk factors (Mountjoy et al., 2010, 2015; Schwellnus proteomic data, in combination with the self-reported training et al., 2016; Soligard et al., 2016, 2017; Drew et al., 2017, distress and illness data, would enhance the ability to identify 2018; Prien et al., 2017). The IOC has focused on inappropriate periods of highest risk of inappropriate load management, and management of both internal (e.g., mental stress responses) and potential overreaching in swimmers. external loads (e.g., exercise training and competition workloads) (Schwellnus et al., 2016; Soligard et al., 2017). Load management is a key strategy, according to the IOC, to reduce illness incidence MATERIALS AND METHODS and associated downturns in exercise performance, interruptions in training, missed competitive events, and risk of serious Study Participants medical complications. The IOC recommends monitoring for Thirty-five swimmers were initially recruited to the study (Sprint early signs and symptoms of over-reaching, overtraining, and Specialty n = 14; Middle Distance Specialty n = 17; and Distance illness, and to periodically assess psychological stresses using available instruments. Multiple studies have established that a mismatch exists TABLE 1 | Characteristics of the n = 32 study participants (mean ± SE). between the exertion stress perceived by athletes and that Variable Male (n = 19) Female (n = 13) intended by coaches (Brink et al., 2014, 2017). For this and other reasons, there is growing interest in the application of Age (years) 19.9 ± 0.4 20.0 ± 0.3 practical and efficacious blood and urine measures that can be Height (m) 1.83 ± 0.02 1.69 ± 0.02 used to improve the detection of overreaching and overtraining Weight (kg) 75.6 ± 3.0 64.8 ± 2.3 in athletes. Unfortunately, reliable biomarkers that are sensitive % Body fat 12.7 ± 1.2 22.6 ± 1.2 to the training load and occur prior to the establishment of VO2max (ml. kg−1 min−1 ) 55.9 ± 1.4 48.1 ± 1.9 overreaching, training distress, and illness have not yet been Second formatting option for Table 1: identified (Meeusen et al., 2013). Subjective measures (i.e., survey Age (years) 19.9 ± 0.4 20.0 ± 0.3 data, rate perceived exertion, and Training Distress Scale) are Height (m) 1.83 ± 0.02 1.69 ± 0.02 currently regarded as superior to physiological measures such Weight (kg) 75.6 ± 3.0 64.8 ± 2.3 as plasma hormones and cytokines, energy homeostasis, and % Body fat 12.7 ± 1.2 22.6 ± 1.2 exercise workload monitoring (Jürimäe et al., 2011; Saw et al., Total Lean Muscle (Kg) 33.4 ± 3.4 21.6 ± 2.5 2016; Joro et al., 2017). VO2max (ml. kg−1 min−1 ) 55.9 ± 1.4 48.1 ± 1.9 Frontiers in Physiology | www.frontiersin.org 2 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers FIGURE 1 | Training distress scale (TDS) total scores (mean ± SE) and upper respiratory tract infection (URTI) prevalence grouped (2–3 weeks running averages) across the 25-week season in n = 30 swimmers. TDS and illness significantly increase above baseline, and remained elevated. Specialty n = 5). Participants voluntarily signed informed et al., 2005). Fingerstick blood samples were then collected using consent forms, and all research procedures were approved by the Volumetric Absorptive Microsampling (VAMS) technology the Queens University of Charlotte Institutional Review Board (Neoteryx, Torrance, CA, United States). Coaches provided (FILE #5-17-BCOH-363). Participants completed baseline testing weekly feedback on each athlete regarding swim performance, within the first two weeks of recruitment. Demographic and signs of training distress, and training intensity (10-point training histories were acquired with questionnaires. Height, Likert scale, with 10 being the most intense). Coaches used body mass, and percent body fat were measured (seca Medical these qualifiers when recording swimmer performance: 1-Below Body Composition Analyzer 514 bioelectrical impedance scale, expectations, 2-fair, 3-average, 4-better than expected, and 5-far Hanover, MD, United States). VO2max was assessed using exceeded expectations. For training distress, the coaches used the Bruce’s treadmill protocol, with oxygen consumption and these qualifiers: 1-no signs, 2-slightly distressed, 3- distressed, 4- ventilation continuously monitored using the Cosmed Fitmate recognizable fatigue and drop in performance, and 5- likely FOR metabolic system (Cosmed, Rome, Italy) (Nieman et al., 2006). with inability to train at expected levels. Coaches were sent the list Swimmers in this study were competing for an NCAA Division of swimmers participating in the study with the above questions II school, and both men’s and women’s swim teams won the in a spreadsheet that they filled out based on the previous week national title the year of data collection (which was the 4th of observations. The same coach assessed the same swimmers to national title consecutively). The average number of “pool” hours reduce inter-rater variability. ranged from 15–18 per week (depending mainly on the specialty sub-group of the swimmers, with distance swimmers clocking Training Distress Scale and Illness more pool hours). The number of weight room hours was not quantified in this study. Monitoring The TDS (Grove et al., 2014) includes 19 questions, and the athletes reported symptoms from the previous week using a 0– Research Design 4 scale. The total TDS score was calculated using the sum of Athletes provided blood samples, illness symptoms, and ratings responses from the 19 questions. The first question of the WURSS of training distress every Monday morning for 19 weeks out of (Barrett et al., 2005) was used to monitor illness severity from the the 25-week season. On Mondays at approximately 5:45 am, the previous week (0 = not sick, 1 = very mild URTI to 7 = severe). athletes reported to the training facility in their normal state (no restrictions on food or water or rest) and completed the Training Proteomics Profiling Distress Scale (TDS) and Wisconsin Upper Respiratory Symptom Dried blood spot (DBS) specimens (approximately 1–2 drops of Survey (WURSS) questionnaires (Grove et al., 2014; Barrett blood, or 40–60 µL of blood/sample) using VAMS (Neoteryx, Frontiers in Physiology | www.frontiersin.org 3 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers FIGURE 2 | (A) Variances of the Principal Components (PC’s) from the PCA analysis of the TDS questionnaire data indicated that ∼40.5% of variance was explained by PC1. (B) PC1 was largely attributed to illness prevalence (URTI) (0 = no illness; >0 = illness). Frontiers in Physiology | www.frontiersin.org 4 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers FIGURE 3 | PC1 for TDS was largely attributed to illness prevalence (URTI). The boxplot shows the differences between TDS scores based on URTI prevalence for PC1 (p < 0.001)1; (0 = no illness; >0 = illness). Torrance, CA, United States) were dried overnight and stored accelerator voltage was set to 4 and the cycle time was set to with desiccant. Proteins from the DBS samples were solubilized 1,500 ms to measure 10 data points per peptide peak. Collision and reduced in 80 µl 8 M Urea, 50 mM AmBiC and 0.1 mM DTT energy and fragmentor were peptide specific and optimized prior for 30 min at 37◦ C. 6 µl of the protein content (∼30 µg) was to measurements. Spectra were analyzed for quality using Skyline transferred to a new 96 wells plate and 1pmol heavy standards (MacCoss Lab Software1 ) with manual validation. was added together with 0.6 µg of trypsin (1:50) in a total volume of 45 µl of 50 mM AmBiC. Tryptic digestion was done Method Development and Limit of Quantitation O/N while shaking at 37◦ C. The next day 5 µl of 10% FA was In order to reliably identify the targeted peptides, heavy labeled added to quench the tryptic digestion. Samples were cleaned synthetic peptides were ordered for every targeted peptide up using C18 reverse phase columns in 96 well plate format (JPT Peptide Technologies, Berlin, Germany). Using Skyline (Waters Sep-Pak , C18, 40 mg) and dried down. All samples R the retention time of each peptide was measured and the were randomized and subsequently measured consecutively by fragmentation of each peptide was optimized by changing the mass spectrometry (MS). MS analyses were performed on a Triple collision energy for every peptide. The MS method was optimized Quad mass spectrometer (Agilent 6460 Triple Quad) coupled to using Masshunter software (Agilent Technologies, Santa Clara, a normal flow LC autosampler (Agilent 1290 Infinity). 20 µg of CA, United States) with the requirement of acquiring at least peptide of each sample was injected and peptides were separated 10 data points across all peptide peaks for accurate quantitation. with reverse phase normal flow LC chromatography. Samples Isolation windows were set to 1 min. As a last step, for every were loaded on a Liquid Chromatography, 2.1 cm × 25 cm peptide a Limit of Quantitation (LOQ) curve was made to C18 2.7 µm 120 Å column (651750-902, AdvanceBio Peptide ascertain that the quantity of peptides measured was within Mapping, Agilent) with a flow rate of 0.2 mL/min (buffer A, the linear range of the curve. Heavy standards were spiked HPLC H2 O, 0.1% FA; buffer B, 100% ACN, 0.1% FA; 40-min in the DBS matrix in 8 different concentrations and every gradient; 0–3 min: 5% buffer B, 3–18 min: 5 to >25% buffer B, concentration was injected three times as technical replicates. 18–26 min: 25 to >29.5% buffer B, 26–30 min: 29.5 to >40% When the coefficient of variation (CV) was greater than 0.2, the buffer B, 30–35 min: 95% buffer B, 35–40 min, 5% buffer B). data at these intensities for a particular peptide were considered Peptides were transferred to the gaseous phase with positive ion electrospray ionization at 5.0 kV. Gas flow was 12 l min−1 , cell 1 https://skyline.ms/ Frontiers in Physiology | www.frontiersin.org 5 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers FIGURE 4 | Generalized Linear Mixed Models (GLMM) representation of 32 GLMM models (y-axis) with the aim to see which of these 5 variables (x-axis) could best explain PC1 variance of the TDS Principal Component Analysis (PCA). Blue and white colors represent no significant relationship with PC1 when a variable was or was not taken into account in the respective model (white and blue, respectively). Red represents when there was a relationship between the variable and PC1 (with darker red representing a stronger association). This figure clearly shows that PC1 was largely attributed to URTI and only to some extent to the Distress score and Group (pink color). Values are in –log(p-value). This GLMM analysis shows that (a) there was a strong correlation between PC1 and URTI (0 = no illness, >0 = illness); (b) there was no significant correlations between PC1 and the coaches’ ratings of distress (Distr_score) or swim performance (Perf_score) in the athletes, the swimmer training group (Group = sprinters, middle-, and long-distance), or sex (male and female). too variable to be used for quantitation purposes and were not then filtered based on their LOQ results; only peptides that had used for further analysis. an intensity in the linear range of the LOQ curve were kept for further analysis. Data was then normalized based on the median Library Creation intensity of the signal of the heavy peptides. Peptides were finally For high throughput data analysis, a library was created based merged into proteins based on their relative median intensity. on the fractionation information of heavy labeled standards (JPT peptide technologies) obtained during method development. The library included the information both for native “light” peptides Statistical Analysis and their “heavy” labeled synthetic counterparts. The format of Principal Component Analysis (PCA) was used from the the library was applicable to the OpenSWATH workflow (Röst FactoMiner R library to analyze TDS scores to find the subspace et al., 2014) used for Data Independent Acquisition (DIA). corresponding to the maximum-variance directions in the original space (Lê et al., 2008). In order to obtain the meaningful Data Processing factors which explained the higher variation, 32 linear mixed The MRM (Multiple Reaction Monitoring) files were processed models were created by combining different variables, with using OpenSWATH software and Skyline. Skyline was used for consideration for a random effect intercept by the athlete. The method development and peptide verification. OpenSWATH was variables included in this analysis were the coaches rating of used for high through-put data MRM signal processing in order distress (Coach_distr) and performance (Coach_perf) for the to identify and quantify peptides. After the first data processing athletes, swimmer group (based on swim race distance), sex, by OpenSWATH, the data was treated as follows: first, peaks were URTI (using binary data, with “0” indicating no illness, and smoothened followed by intensity correction of the fragments “>0” indicating illness during the defined time period), and the according to the ratio of the fragments as was present in the first principal component as a response variable. P-values were library (as defined during method optimization). Peptides were obtained through the ANOVA procedure. Frontiers in Physiology | www.frontiersin.org 6 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers FIGURE 5 | Generalized Linear Mixed Models results of the 12 targeted chronic proteins in the study (y-axis) across the weeks (x-axis). White rectangles mean that there was no significant difference in expression for the specific protein for that week compared to the baseline weeks. The intensity of blue color of the rectangles indicates the significance of the difference in expression of the proteins between those weeks and the baseline values. Values are in –log(p-value). A Generalized Linear Mixed Models (GLMM) approach was term into the loss function forcing some coefficients to be used to model the response of the 12 targeted proteins across zero. Differences between scores from LASSO output were the weeks compared to the baseline week. In the models, the compared using the Kruskal–Wallis non-parametric rank sum athlete was used as the random effect and sex as the fixed test (Kruskal and Wallis, 1952). effect. Pairwise comparisons between each time-by-condition level were calculated using the glht function from multcomp R package (Hothorn et al., 2008). The Tukey correction for RESULTS multiple comparisons was applied to adjust the significance level in this study. Of the thirty-five swimmers recruited to the study, 32 swimmers The (LDA analysis was used to evaluate the separation completed the protocol. See Table 1 for subject characteristics. between groups based on different time periods: baseline, Figure 1 shows the TDS scores (mean ± SE) and URTI and when athletes were taking exams, competing in meets, prevalence across the 25-week season in the swimmers. The data or engaged in both exams and competitive meets. In the were grouped in 2–3 weeks segments to facilitate interpretation beginning, LDA was focused only on meta-data (TDS, and and analysis of the data with respect to training blocks and URTI = 0/>0) and then secondly, the analysis focused academic periods. TDS scores varied between athletes, and on the additional effect of the protein data. The LDA were highest during the early fall conditioning ramp up analysis included leave-by-one CV and calculated p-values period (8.9 ± 1.6 at baseline to a peak of 22.6 ± 2.0). The from the Kruskal–Wallis non-parametric rank sum test percent of athletes reporting illness was high throughout the (Kruskal and Wallis, 1952). season (50.0–78.1%). The LASSO (Least Absolute Shrinkage and Selection Analysis of TDS data using PCA revealed that 40.8% of the Operation) logistic regression model (Tibshirani, 1996) variance was explained by PC1 (Figure 2A). PC1 was largely was used to evaluate differences between time segments attributed to URTI prevalence (Figure 2B). Boxplots in Figure 3 (baseline versus exams, meets, and meets and exams, and show the differences in TDS scores based on URTI prevalence also meets versus meets and exams). LASSO is a powerful for PC1. The centered PC1 TDS scores for the athletes reporting regularization technique and incorporates an L1- penalization illness and no illness were different across the season (P < 0.001). Frontiers in Physiology | www.frontiersin.org 7 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers FIGURE 6 | (A) Linear Discriminant Analysis (LDA) analysis taking into account TDS, illness, and protein data. Red = baseline weeks; green = exam weeks; blue = swim meet weeks; purple = exam and swim meet weeks. This analysis without the protein expression data supported a separation of baseline weeks from all other weeks (p < 0.001, Kruskal–Wallis). (B) The same LDA analysis after the inclusion of the 12 selected proteins. This analysis showed an improved, distinct separation of the baseline weeks from all other time periods (p < 0.001, Kruskal–Wallis). Frontiers in Physiology | www.frontiersin.org 8 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers FIGURE 7 | Logistic LASSO regression including the TDS total score, coaches’ ratings of swimmer’s distress and performance, and 12 selected blood proteins showed a significant separation when comparing exams and meets with exams time segments (p-value < 0.001). The regression analysis between these time periods was not significant when the protein data were removed. TABLE 2 | The 12 targeted proteins measured in this study based on the data from Nieman et al. (2018). UniProt protein Protein name Function P35542 Serum amyloid A-4 protein Major acute phase reactant; cell chemotaxis P05164 Myeloperoxidase Granulocyte microbicidal activity; production of hypochlorous acid P07360 Complement component C8 gamma chain Part of membrane attack complex; forms pores in target cells P0C0L5 Complement C4B Non-enzymatic component C3, C5 convertases; complement activation; inflammation P05155 Plasma protease C1 inhibitor Crucial role in complement activation. Q14624 Inter-alpha-trypsin inhibitor heavy chain H4 Acute-phase protein involved in trauma inflammatory response P19652 Alpha-1-acid glycoprotein 2 Transport protein; modulates immune; acute-phase; inflammation P10643 Complement component C7 Part of membrane attack complex; forms pores in target cells P02765 Alpha-2-HS-glycoprotein Promotes endocytosis; acute-phase response; phagocytosis; bone mineral P01834 Ig kappa chain C region Antigen and Ig receptor binding; complement activation; innate; phagocytosis P08185 Corticosteroid-binding globulin Major transport protein for glucocorticoids and progestins P35754 Glutaredoxin-1 Glutathione activity; cell redox homeostasis The GLMM representation of 32 analysis models is shown selected weeks. To improve interpretation of these data, LDA in Figure 4. The goal of this analysis was to explain analysis was conducted with TDS and URTI data included and PC1 variance of the TDS PCA. This analysis showed that sorted into periods when the swimmers were taking exams there was a strong correlation between URTI and PC1 with normal training (exams), engaging in competitive meets (red colors). The coaches’ ratings of swim performance and (meets), and when the swimmers were taking both exams swimmer’s distress, swimmer’s sex, and racing distance were not and competing in meets (meet-exam). This LDA analysis correlated with PC1. was conducted both with and without the proteomics data Generalized Linear Mixed Models analysis of protein (Figures 6A,B). This analysis showed a separation of the expression across the season in the swimmers is shown baseline weeks from weeks with exams, meets, or both exams in Figure 5. The data support protein expression during and meets (p < 0.001, Kruskal–Wallis). Adding the panel of Frontiers in Physiology | www.frontiersin.org 9 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers protein expression data enhanced the separation between these in signaling pathways related to phagocyte migration and groups of weeks (p < 0.001, Kruskal–Wallis). Figure 7 depicts inflammation (Ye and Sun, 2015). Although more human studies results from the logistic LASSO regression that included the are needed to improve practical applications in the athletic TDS total score, coaches’ ratings of swimmer’s distress and setting, SAA, one of the proteins in the targeted protein panel performance, and 12 selected blood proteins. This analysis (Table 2), has been used in studies of race horses and dogs showed a separation of weeks when exams were taken from as an indicator of exercise-induced muscle damage and poor weeks when both competitive swim meets and exams were performance (Cywinska et al., 2010, 2013; Casella et al., 2013; engaged in (p-value < 0.001). The regression analysis between Valle et al., 2015). these time periods was not significant when the protein The data of the current study supports the use of this targeted data were removed. protein panel in combination with ratings of training distress when monitoring athletes during a competitive training season. The chief limitation of the current study was the lack of a DISCUSSION suitable control group, and future research will help define DBS protein levels that when combined with TDS scores predict This study tracked changes in a targeted group of proteins, overtraining. TDS, illness, and proteomics data can be used to recently identified as potential biomarkers of overreaching in support decisions by the coaches or performance monitoring athletes (Nieman et al., 2018), in high level swimmers over team regarding alleviation of both internal and external loads the course of the entire season. Indeed, combining proteomic for the athlete. A unique finding in this study was the apparent data with self-reported training distress and illness, allowed for importance of the internal stressors to the University level identification of periods of high physiologic stress and potential athlete, such as academic exam periods, psychological pressure to overreaching in swimmers. perform athletically, and other personal stressors. These internal The swimmers in this study experienced high levels of distress stressors contribute to the physiologic response in the body to and URTI (prevalence of 50.0–78.1% during each 2 to 3-week external training loads, and future research should investigate segment) throughout the 25-week season, confirming reports strategies to mitigate the negative response of both internal and from other investigators (Hellard et al., 2015; Collette et al., external stressors of athletes. 2018). Hellard et al. (2015) followed 28 elite swimmers for 4 years and showed that the odds of URTI were 50–70% higher during intensive training periods. Studies and comprehensive CONCLUSION reviews generally support the relationship between training loads This study showed that training distress was strongly related to and illness, especially during periods of intensification and URTI prevalence in collegiate swimmers, and that proteomics competition (Fricker et al., 2005; Gleeson et al., 2013; Aubry et al., data added strength to this relationship, especially during high 2014; Drew and Finch, 2016; Jones et al., 2017). stress periods when the athletes were involved with exams and/or The physiologic and psychologic response to both internal competitive swim meets. Coaches feedback regarding training and external stress is highly individualized in swimmers (Collette distress and level of swim performance did not correlate well et al., 2018). This variance underscores the importance of with the swimmer’s TDS scores, confirming the findings of adding physiologic biomarkers to aid in the identification of other studies (Brink et al., 2014, 2017). One potential cause overreaching and overtraining. In a previous study conducted by for the disconnect between athlete and coach is that periods of our research group, a targeted panel of proteins was identified physiologic stress are related at least in part to factors outside that could be utilized and validated in future investigations the training program. According to the NCAA, 73% of college of overreaching and overtraining (Nieman et al., 2018). These athletes believe their coach cares about their overall well-being proteins can be analyzed in fingerprick DBS samples, improving (Paskus, 2016). Despite this, 30% of athletes report they have the potential practical application of this technology to athletic been overwhelmed during the past month, and desire more settings compared to full blood draws. The targeted protein focus by the coaches on their education, sleep, and nutrition. panel met the requirement that changes in response to acute Division II NCAA athletes report spending 32.5 h/week on exercise should be distinguishable from chronic changes, and athletics, and up to 38.5 h/week on academics (Paskus, 2016). be relatively easy to collect and measure (Meeusen et al., Coaches often do not account for all of the stresses experience 2013). Our analysis found that most of these targeted proteins by their athletes, including academic, social, financial, and living were involved in the immune defense response including the challenges. Further research using proteomics data as an objective acute phase response, complement activation, and humoral measure will refine strategies to reduce training stress and illness responses mediated by circulating immunoglobulins (Nieman prevalence among elite athletes, and improve coach’s perceptions et al., 2018). The acute phase response is a systemic reaction of training-related distress. to prolonged exercise stress, and involves the production of many proteins including serum amyloid A (SAA), complement and transport proteins, antiproteases, and those involved with DATA AVAILABILITY STATEMENT the coagulation and fibrinolytic system (Gabay and Kushner, 1999; Balfoussia et al., 2014). Liver production of SAA rises The datasets generated for this study are available on request to strongly during the acute phase response, and is involved the corresponding author. Frontiers in Physiology | www.frontiersin.org 10 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers ETHICS STATEMENT FUNDING The studies involving human participants were reviewed Internal funds were used to cover the costs associated and approved by Institutional Review Board, Queens with this study. University of Charlotte. The patients/participants provided their written informed consent to participate in this study. ACKNOWLEDGMENTS We would like to thank the coaches, swimmers, and athletic AUTHOR CONTRIBUTIONS directors at Queens University of Charlotte for their participation with this research study. We also express gratitude to the AK, DN, LZ, AG, and AP designed the study and following exercise and sport science student interns who assisted drafted the manuscript. AK collected the data. AG in data collection: Kyrie Dobson, McKenzie Stevens, Lola and AP analyzed the blood samples. AK, DN, LZ, Bulatova, Sydney Ditto, Drew Breese, Leanne Bryant, Erin AG, and AP analyzed the data. All authors are Iknayan, Deja Lewis-Donnerson, Liam Doyle, and Michael James. responsible for revising the intellectual content of the We acknowledge the assistance of Alexander Rakitko of the manuscript, and reading and approving the final version Lomonosov Moscow State University in Russia for assistance of the manuscript. with the data analysis. REFERENCES Fricker, P. A., Pyne, D. B., Saunders, P. U., Cox, A. J., Gleeson, M., and Telford, R. D. (2005). Influence of training loads on patterns of illness in elite distance Aubry, A., Hausswirth, C., Louis, J., Coutts, A. J., and Meur, Y. L. E. (2014). runners. Clin. J. Sport Med. 15, 244–250. Functional overreaching: the key to peak performance during the taper? Med. Gabay, C., and Kushner, I. (1999). Acute-phase proteins and other systemic Sci. Sports Exerc. 46, 1769–1777. doi: 10.1249/mss.0000000000000301 responses to inflammation. N. Engl. J. Med. 340, 448–454. doi: 10.1056/ Balfoussia, E., Skenderi, K., Tsironi, M., Anagnostopoulos, A. K., Parthimos, N., nejm199902113400607 Vougas, K., et al. (2014). A proteomic study of plasma protein changes under Gleeson, M., Bishop, N., Oliveira, M., and Tauler, P. (2013). Influence of training extreme physical stress. J. Proteomics 98, 1–14. doi: 10.1016/j.jprot.2013.12.004 load on upper respiratory tract infection incidence and antigen-stimulated Barrett, B., Brown, R., Mundt, M., Safdar, N., Dye, L., Maberry, R., et al. (2005). cytokine production. Scand. J. Med. Sci. Sports 23, 451–457. doi: 10.1111/j.1600- The Wisconsin Upper Respiratory Symptom Survey is responsive, reliable, and 0838.2011.01422.x valid. J. Clin. Epidemiol. 58, 609–617. doi: 10.1016/j.jclinepi.2004.11.019 Gleeson, M., McDonald, W. A., Cripps, A. W., Pyne, D. B., Clancy, R. L., and Brink, M. S., Frencken, W. G. P., Jordet, G., and Lemmink, K. A. (2014). Coaches’ Fricker, P. A. (1995). The effect on immunity of long-term intensive training in and players’ perceptions of training dose: not a perfect match. Int. J. Sports elite swimmers. Clin. Exp. Immunol. 102, 210–216. doi: 10.1111/j.1365-2249. Physiol. Perform. 9, 497–502. doi: 10.1123/ijspp.2013-0009 1995.tb06658.x Brink, M. S., Kersten, A. W., and Frencken, W. G. P. (2017). Understanding Gleeson, M., McDonald, W. A., Pyne, D. B., Clancy, R. L., Cripps, A. W., Francis, the mismatch between coaches’ and players’ perceptions of exertion. J. L., et al. (2000). Immune status and respiratory illness for elite swimmers Int. J. Sports Physiol. Perform. 12, 562–568. doi: 10.1123/ijspp.2016- during a 12-week training cycle. Int. J. Sports Med. 21, 302–307. doi: 10.1055/ 0215 s-2000-313 Casella, S., Fazio, F., Russo, C., Giudice, E., and Piccione, G. (2013). Acute phase Grove, J. R., Main, L. C., Partridge, K., Bishop, D. J., Russell, S., Shepherdson, proteins response in hunting dogs. J. Vet. Diagn. Investig. 25, 577–580. doi: A., et al. (2014). Training distress and performance readiness: laboratory and 10.1177/1040638713495851 field validation of a brief self-report measure. Scand. J. Med. Sci. Sports 24, Collette, R., Kellmann, M., Ferrauti, A., Meyer, T., and Pfeiffer, M. (2018). e483–e490. doi: 10.1111/sms.12214 Relation between training load and recovery-stress state in high-performance Hellard, P., Avalos, M., Guimaraes, F., Toussaint, J. F., and Pyne, D. B. (2015). swimming. Front. Physiol. 9:845. doi: 10.3389/fphys.2018.00845 Training-related risk of common illnesses in elite swimmers over a 4-yr period. Cywinska, A., Gorecka, R., Szarska, E., Witkowski, L., Dziekan, P., and Med. Sci. Sports Exerc. 47, 698–707. doi: 10.1249/MSS.0000000000000461 Schollenberger, A. (2010). Serum amyloid A level as a potential indicator of the Hothorn, T., Bretz, F., and Westfall, P. (2008). Simultaneous inference in general status of endurance horses. Equine Vet. J. Suppl. 38, 23–27. doi: 10.1111/j.2042- parametric models. Biomater. J. 50, 346–363. doi: 10.1002/bimj.200810425 3306.2010.00280.x Jones, C. M., Griffiths, P. C., and Mellalieu, S. D. (2017). Training load and fatigue Cywinska, A., Witkowski, L., Szarska, E., Schollenberger, A., and Winnicka, marker associations with injury and illness: a systematic review of longitudinal A. (2013). Serum amyloid A (SAA) concentration after training sessions in studies. Sports Med. 47, 943–974. doi: 10.1007/s40279-016-0619-5 Arabian race and endurance horses. BMC Vet. Res. 9:91. doi: 10.1186/1746- Joro, R., Uusitalo, A., DeRuisseau, K. C., and Atalay, M. (2017). Changes in 6148-9-91 cytokines, leptin, and IGF-1 levels in overtrained athletes during a prolonged Drew, M., Vlahovich, N., Hughes, D., Appaneal, R., Burke, L. M., Lundy, B., et al. recovery phase: a case-control study. J. Sports Sci. 35, 2342–2349. doi: 10.1080/ (2018). Prevalence of illness, poor mental health and sleep quality and low 02640414.2016.1266379 energy availability prior to the 2016 summer olympic games. Br. J. Sports Med. Jürimäe, J., Mäestu, J., Jürimäe, T., Mangus, B., and von Duvillard, S. P. (2011). 52, 47–53. doi: 10.1136/bjsports-2017-098208 Peripheral signals of energy homeostasis as possible markers of training stress in Drew, M. K., and Finch, C. F. (2016). The relationship between athletes: a review. Metabolism 60, 335–350. doi: 10.1016/j.metabol.2010.02.009 training load and injury, illness and soreness: a systematic and Knab, A. M., Nieman, D. C., Gillitt, N. D., Shanely, R. A., Cialdella-Kam, L., literature review. Sports Med. 46, 861–883. doi: 10.1007/s40279-015- Henson, D. A., et al. (2013). Effects of a flavonoid-rich juice on inflammation, 0459-8 oxidative stress, and immunity in elite swimmers: a metabolomics-based Drew, M. K., Vlahovich, N., Hughes, D., Appaneal, R., Peterson, K., Burke, L., et al. approach. Int. J. Sport Nutr. Exerc. Metab. 23, 150–160. doi: 10.1123/ijsnem. (2017). A multifactorial evaluation of illness risk factors in athletes preparing 23.2.150 for the summer olympic games. J. Sci. Med. Sport 20, 745–750. doi: 10.1016/j. Kruskal, W. H., and Wallis, W. A. (1952). Use of ranks in one-criterion variance jsams.2017.02.010 analysis. J. Am. Stat. Assoc. 47, 583–621. doi: 10.1080/01621459.1952.10483441 Frontiers in Physiology | www.frontiersin.org 11 May 2020 | Volume 11 | Article 373 Knab et al. Proteomics Monitoring in Swimmers Lê, S., Josse, J., and Husson, F. (2008). FactoMineR: an R package for multivariate Saw, A. E., Main, L. C., and Gastin, P. B. (2016). Monitoring the athlete training analysis. J. Stat. Softw. 25, 1–18. response: subjective self-reported measures trump commonly used objective Meeusen, R., Duclos, M., Foster, C., Fry, A., Gleeson, M., Nieman, D. C., measures: a systematic review. Br. J. Sports Med. 50, 281–291. doi: 10.1136/ et al. (2013). European College of Sport Science; American College of Sports bjsports-2015-094758 Medicine. Prevention, diagnosis, and treatment of the overtraining syndrome: Schwellnus, M., Soligard, T., Alonso, J. M., Bahr, R., Clarsen, B., Dijkstra, joint consensus statement of the European College of Sport Science and the H. P., et al. (2016). How much is too much? (Part 2) international American College of Sports Medicine. Med. Sci. Sports Exerc. 45, 186–205. olympic committee consensus statement on load in sport and risk of Mountjoy, M., Junge, A., Alonso, J. M., Engebretsen, L., Dragan, I., Gerrard, illness. Br. J. Sports Med. 50, 1043–1052. doi: 10.1136/bjsports-2016- D., et al. (2010). Sports injuries and illnesses in the 2009 FINA world 096572 championships (Aquatics). Br. J. Sports Med. 44, 522–527. doi: 10.1136/bjsm. Soligard, T., Schwellnus, M., Alonso, J. M., Bahr, R., Clarsen, B., Dijkstra, 2010.071720 H. P., et al. (2016). How much is too much? (Part 1) International Mountjoy, M., Junge, A., Benjamen, S., Boyd, K., Diop, M., Gerrard, D., et al. Olympic Committee consensus statement on load in sport and risk (2015). Competing with injuries: injuries prior to and during the 15th FINA of injury. Br. J. Sports Med. 50, 1030–1041. doi: 10.1136/bjsports-2016- world championships 2013 (aquatics). Br. J. Sports Med. 49, 37–43. doi: 10.1136/ 096581 bjsports-2014-093991 Soligard, T., Steffen, K., Palmer, D., Alonso, J. M., Bahr, R., Lopes, A. D., et al. Nieman, D. C. (2000). Is infection risk linked to exercise workload? Med. Sci. Sports (2017). Sports injury and illness incidence in the Rio de Janeiro 2016 Olympic Exerc. 32(7 Suppl. 7), S406–S411. Summer Games: a prospective study of 11274 athletes from 207 countries. Br. J. Nieman, D. C., Austin, M. D., Benezra, L., Pearce, S., McInnis, T., Unick, J., et al. Sports Med. 51, 1265–1271. doi: 10.1136/bjsports-2017-097956 (2006). Validation of Cosmed’s Fitmate in measuring oxygen consumption and Tibshirani, R. (1996). Regression shrinkage and selection via the Lasso. J. R. Stat. estimating resting metabolic rate. Res. Sports Med. 14, 89–96. doi: 10.1080/ Soc. Series B Stat. Methodol. 73, 267–288. 15438620600651512 Valle, E., Zanatta, R., Odetti, P., Traverso, N., Furfaro, A., Bergero, D., et al. Nieman, D. C., Groen, A. J., Pugachev, A., and Vacca, G. (2018). Detection of (2015). Effects of competition on acute phase proteins and lymphocyte functional overreaching in endurance athletes using proteomics. Proteomes subpopulations—Oxidative stress markers in eventing horses. J. Anim. Physiol. 6:33. doi: 10.3390/proteomes6030033 Anim. Nutr. 99, 856–863. doi: 10.1111/jpn.12289 Nieman, D. C., and Wentz, L. M. (2019). The compelling link between physical Ye, R. D., and Sun, L. (2015). Emerging functions of serum amyloid A in activity and the body’s defense system. J. Sport Health Sci. 8, 201–217. doi: inflammation. J. Leukoc Biol. 98, 923–929. doi: 10.1189/jlb.3VMR0315-080R 10.1016/j.jshs.2018.09.009 Paskus, (2016). NCAA GOALS Study of the Student-Athlete Experience. Initial Conflict of Interest: AG and AP are both founders of the company ProteiQ Summary of Findings. Available at: http://www.ncaa.org/sites/default/files/ Biosciences GmbH in Berlin, Germany. LZ was under contract with ProteiQ GOALS_2015_summary_jan2016_final_20160627.pdf (accessed September 25, Biosciences GmbH at the time of statistical analysis for this paper. 2017). Prien, A., Mountjoy, M., Miller, J., Boyd, K., van den Hoogenband, C., Gerrard, The remaining authors declare that the research was conducted in the absence of D., et al. (2017). Injury and illness in aquatic sport: how high is the risk? A any commercial or financial relationships that could be construed as a potential comparison of results from three FINA world championships. Br. J. Sports Med. conflict of interest. 51, 277–282. doi: 10.1136/bjsports-2016-096075 Rodríguez-Suárez, E., and Whetton, A. D. (2013). The application of quantification Copyright © 2020 Knab, Nieman, Zingaretti, Groen and Pugachev. This is an open- techniques in proteomics for biomedical research. Mass Spectr. Rev. 32, 1–26. access article distributed under the terms of the Creative Commons Attribution doi: 10.1002/mas.21347 License (CC BY). The use, distribution or reproduction in other forums is permitted, Röst, H. L., Rosenberger, G., Navarro, P., Gillet, L., Miladinović, S. M., Schubert, provided the original author(s) and the copyright owner(s) are credited and that the O. T., et al. (2014). OpenSWATH enables automated, targeted analysis of data- original publication in this journal is cited, in accordance with accepted academic independent acquisition MS data. Nat. Biotechnol. 32, 219–223. doi: 10.1038/ practice. No use, distribution or reproduction is permitted which does not comply nbt.2841 with these terms. Frontiers in Physiology | www.frontiersin.org 12 May 2020 | Volume 11 | Article 373