Participants. Thirteen trained (at least Tier 2; McKay et al. (2021)), non-heat acclimated, endurance runners (11 male, 2 female; mean ± SD: age 31 ± 8 y, stature 1.75 ± 0.06 m, body mass 71.1 ± 10.5 kg, \(\mathop }\limits^\)O2peak 60 ± 6 mL/kg/min) participated in this study, which received ethical approval from the Loughborough University Ethics Approvals (human participants) Sub-Committee (Reference: 2021-4673-4374). Initially, fifteen participants were recruited, but two withdrew following the first experimental visit due to injury unrelated to the study. Participants completed a health screening questionnaire and provided verbal and written informed consent before participating. Participants completed four trials: a preliminary visit, familiarisation trial and two experimental trials at the same time of day (standardised within-participant) in a double blind, placebo-controlled, randomised manner. The familiarisation trial and first experimental trial were separated by ≥ 7 days, and the experimental trials were separated by ≥ 3 weeks, allowing for a 2-week washout period (Ashley et al. 2018) and a further week of supplementation before the second experimental trial. For female participants, trials were separated by the same minimum time frames, but experimental trials also occurred during the same phase of the menstrual or contraception pill cycle (identified via participant confirmation).

Preliminary testing. During the first visit, body mass and stature were measured. Running peak oxygen uptake (\(\mathop }\limits^\)O2peak) was determined using a motorised treadmill (HP Cosmos, Nussdorf, Germany). Participants completed four, 4-min stages at progressively increasing speeds until a heart rate of  >160 beat/min was obtained. Expired gas was collected into a Douglas bag during the final min of each stage and analysed for oxygen and carbon dioxide concentrations (Servomex 1400 Gas Analyser, Servomex, Crowborough, UK), volume (Harvard dry gas meter, Harvard Apparatus Ltd, Edenbridge, UK) and temperature (RS Pro digital thermometer, RS Components, Corby, UK). Ambient air was collected concurrently with expired gases to correct \(\mathop }\limits^\)O2 and \(\mathop }\limits^\)CO2 (Betts and Thompson 2012). Following a self-selected rest period (to ensure sufficient recovery), participants ran to volitional exhaustion via a ramp protocol, where speed of the final sub-maximal stage was maintained, with incline increasing by 1% each min until volitional exhaustion. A final expired gas sample was collected during the final min of exercise. These data were then extrapolated using linear regression to determine the speed equivalent to 60–65% of \(\mathop }\limits^\)O2peak for use in subsequent trials. Following a short rest period (~ 5–10 min), participants practiced the 3 km time trial. During the familiarisation trial, participants completed a replication of the experimental trials (outlined below) including all measures except prior supplementation.

Pretrial standardisation. In the 24 h preceding their first experimental trial, participants recorded their dietary intake and physical activity and replicated these patterns prior to the second experimental trial. In the 24 h before each trial, participants were asked to refrain from vigorous exercise and alcohol consumption, and were asked to consume a minimum of 40 mL/kg body mass of fluid (any fluid excluding alcoholic drinks) to help ensure euhydration upon arrival to the laboratory (Minshull and James 2013). On the day of testing, participants consumed a standardised pre-trial meal providing 1 g carbohydrate/kg body mass and 7 mL/kg body mass of fluid consisting of cereal bars, orange juice (Tesco PLC, Welwyn Garden City, UK) and tap water 45 min before arrival at the laboratory.

Supplementation procedures. Experimental trials were preceded by six consecutive days of supplementation, with the seventh and final dosage taken on the day of experimental trials. All supplementation consisted of 6 g/day L-citrulline (NOW Foods, Bloomingdale, IL, USA) (CIT trial) or 6 g/day maltodextrin (MyProtein, Cheshire, UK) as a placebo (PLA trial). Each daily supplement dose was provided in 12, size 00 vegan capsules (Bulk, Essex, UK). Participants were instructed to consume daily supplements at a similar time of day as their scheduled trial time (outside of mealtimes), with ~ 300 mL of water and within a 5 min period.

During experimental trials, participants consumed the final supplement dose with 300 mL diluted sugar-free squash, with 6 g (CIT) or 0 g (PLA) maltodextrin powder to match acute carbohydrate intake between trials. Experimental supplementation and drinks were administered in a double-blind manner, where an investigator not involved in data collection randomised the supplements and made up the drinks. The timing of this supplementation was in line with previous pharmacokinetic data (Schwedhelm et al. 2008) indicating peak plasma L-arginine ~ 90–100 min after 6 g L-citrulline oral supplementation. The current protocol was designed to illicit peak plasma L-arginine at the beginning of the time trial (Table 1).

Experimental trials. Participants ingested a radiotelemetry pill (BodyCap, e-Celsius® Performance, Hérouville Saint-Clair, France) 8–10 h prior to arrival to measure body core temperature. Upon arrival, participants provided a total void urine sample, before nude body mass was recorded. Participants were fitted with a wireless heart rate monitor (Polar M400, Polar, Kemple, Finland) and four wireless skin thermistors (iButton DS1922L Thermochron Data Logger; Berkshire; UK) on the right chest, triceps, thigh and calf, for measurement of heart rate and skin temperature respectively, with weighted mean skin temperature calculated (Ramanathan 1964). Thereafter, participants assumed an upright seated position, and after 20 min, venous blood was collected from an antecubital/forearm vein and core and skin temperature recorded. Participants then consumed their final supplement dose and 40 min later, a venous blood sample and urine sample were collected, before nude body mass was measured. Participants then entered the environmental chamber (32 °C and 50% relative humidity) and completed 50 min continuous running at 60–65% \(\mathop }\limits^\)O2peak (preload). During the preload of both trials, participants consumed a small bolus of water (0.5 mL/kg body mass) at 25 and 50 min (total 36 ± 5 mL) to minimise discomfort and to allow hypohydration to accrue during exercise. Upon completion of the preload, participants sat for an immediate venous blood sample, before nude body mass was measured. Participants then completed a 3 km time trial. For the time trial, the treadmill was stationary, and the investigator counted down from 3 (3, 2, 1, GO!), with participants able to control treadmill speed via a controller proximal to their right side. Participants were told that this was ‘a performance test’ and were instructed to ‘complete the 3 km as quickly as possible’ and were blinded to all feedback, except distance completed. The same investigator sat behind the treadmill and verbally notified the participants at each kilometre using standardised wording, with no further interaction between investigator and participant. Upon completion of the time trial, participants exited the chamber and towel dried, before nude body mass was measured, and a urine sample collected. Finally, participants sat for 20 min before a venous blood sample was collected.

Measures. Heart rate, core and skin temperature were recorded at baseline, 40 min following supplementation (pre-preload), every 10 min throughout the preload, and every 1 km of the time trial. Expired gas samples were collected at 24–25 and 49–50 min of the preload. Rating of perceived exertion (RPE; (Borg 1982)), GI comfort (Jeukendrup et al. 2000) and thermal sensation (RTS; −10 to + 10 scale; (Lee et al. 2008)) were recorded at baseline, 25 and 50 min of the preload and at the end of the time trial. Local sweat rate was measured via placement of a technical absorbent patch on the lower back (Tegaderm Plus; 3M Health-care, Loughborough, U.K) for the final 10 min of the preload. Sweat patch area was measured, and patches weighed to the nearest 0.0001 g (Sartorius, Germany) pre and post application to determine local sweat rate (Morris et al. 2013).

Sample analyses. For each blood sample, 11 mL of blood was drawn, of which 1 mL was dispensed into a tube containing K2 EDTA (1.75 mg/mL; Teklab, Durham, UK) and used to measure haemoglobin concentration via the cyanmethemoglobin method in duplicate and haematocrit via microcentrifugation in triplicate. These values were used to determine changes in blood, red cell, and plasma volume relative to baseline (Dill and Costill 1974). A further 10 mL of blood was dispensed into Lithium Heparin tubes (0.25 mg/mL; Sarstedt AG & CO., Nümbrecht, Germany), centrifuged (2500 g, 20 min, 4 °C) and plasma was stored at −80 °C until analysed for osmolality via freezing-point depression (Gonotec Osmomat 030 Cryoscopic Osmometer; Gonotec, Germany; CV 0.3%), intestinal fatty acid binding protein (I-FABP) via enzyme linked immunosorbent assay (ELISA; R&D Systems, Minneapolis, USA; CV 8.3%), and L-citrulline and L-arginine concentrations via liquid chromatography-tandem mass spectrometry using an adapted protocol previously validated by Shin et al. (2015) [(CV 4.0% L-arginine and 7.5% L-citrulline); see supplementary material for full method].

Statistical analyses. Sample size was determined using G*Power (Version 3.1.9.2) and previous data from our laboratory for the same 3 km performance test (Funnell et al. 2023). It was estimated that nine participants would be sufficient to detect a 5% difference in 3 km running time trial performance with a statistical power of 0.8, and 13 participants with a statistical power of 0.95. Data were analysed using SPSS (version 28; SPSS) and were checked for normality of distribution using a Shapiro–Wilk test. Data containing two factors were analysed using a two-way (trial*time) repeated measures ANOVA. If the assumption of sphericity was violated, the Greenhouse–Geisser estimate was used to correct the degrees of freedom. If significant, interaction and trial effects were followed by a post hoc paired t-test or Wilcoxon signed rank test, depending on normality of data. Familywise error was controlled using a Holm–Bonferroni correction. Data containing one factor (baseline data, time trial performance, local sweat rate) were analysed via paired t-test or Wilcoxon signed rank test, as appropriate. Hedge’s G effect sizes (g) were calculated and interpreted as small (0.2–0.49), medium (0.5–0.79) and large (≥ 0.8). Data sets were accepted as statistically different when P ≤ 0.05. All data were screened for outliers, with one outlier observed for L-citrulline and L-arginine according to the ± 3 MAD method (Leys et al. 2013). Plasma L-citrulline values during exercise for the outlier were > 21 standard deviations above the mean of the other 12 participants, therefore, this was removed from the results. Data are presented as mean ± standard deviation, unless otherwise specified.