Exogenous ketosis elevates circulating erythropoietin and stimulates muscular angiogenesis during endurance training overload

De novo capillarization is a primary muscular adaptation to endurance exercise training and is crucial to improving performance. Excess training load, however, impedes such beneficial adaptations, yet we recently demonstrated that such downregulation may be counteracted by ketone ester ingestion (KE) post‐exercise. Therefore, we investigated whether KE could increase pro‐angiogenic factors and thereby stimulate muscular angiogenesis during a 3‐week endurance training‐overload period involving 10 training sessions/week in healthy, male volunteers. Subjects received either 25 g of a ketone ester (KE, n = 9) or a control drink (CON, n = 9) immediately after each training session and before sleep. In KE, but not in CON, the training intervention increased the number of capillary contacts and the capillary‐to‐fibre perimeter exchange index by 44% and 42%, respectively. Furthermore, KE also substantially increased vascular endothelial growth factor (VEGF) and endothelial nitric oxide synthase (eNOS) expression both at the protein and at the mRNA level. Serum erythropoietin concentration was concomitantly increased by 26%. Conversely, in CON the training intervention increased only the protein content of eNOS. These data indicate that intermittent exogenous ketosis during endurance overload training stimulates muscular angiogenesis. This likely resulted from a direct stimulation of muscle angiogenesis, which may be at least partly due to stimulation of erythropoietin secretion and elevated VEGF activity, and/or an inhibition of the suppressive effect of overload training on the normal angiogenic response to training. This study provides novel evidence to support the potential of exogenous ketosis to benefit endurance training‐induced muscular adaptation.


Introduction
Enhanced skeletal muscle vascularization is an early adaptive response to exercise training, and improves the potential for oxygen diffusion, nutrient uptake and metabolite clearance to and from skeletal muscles (Hudlicka et al., 1992). The increase in muscle capillarization usually precedes or parallels exercise-induced adaptations at the myofibre level, such as a shift from type IIb/IIx to type IIa fibres (Waters et al., 2004) and increased oxidative enzyme activity (Andersen & Henriksson, 1977;Murias et al., 2011). Hence it has been postulated that angiogenesis is required for these latter adaptations to occur (Škorjanc et al., 1998). The angiogenic response to endurance training clearly is an important mechanism involved in the development of endurance performance . But capillarization also plays a pivotal role in maintenance of metabolic health. From a clinical perspective, stimulation of de novo angiogenesis also significantly contributes to the beneficial effects of exercise training in the prevention and treatment of various diseases such as peripheral arterial disease, diabetes and chronic obstructive pulmonary disease (Kissane & Egginton, 2019). Collectively, the above findings clearly indicate that the angiogenic response to exercise is highly important from both a clinical and an exercise performance perspective.
Exercise training induces beneficial muscular adaptations, indeed, yet excessive exercise may also result in so-called over-reaching, and eventually overtraining in a later stage (Bellinger et al., 2019;Ferraresso et al., 2012;Meeusen et al., 2013). Recent studies indicated that training overload and (too) intense exercise can either impair skeletal muscle adaptation or at least deliver a weak stimulus for beneficial adaptations (Bellinger et al., 2019;Hoier et al., 2013). In this regard, Bellinger et al. (2019), demonstrated that 3 weeks of high-volume overload training improved muscular oxidative capacity only in those subjects who did not develop symptoms of over-reaching, but not in the over-reached. This suggests that timely development of adequate muscular responses is important in muscular adaptation to training.
Recently, we demonstrated that post-exercise and pre-sleep ketone ester ((R)-3-hydroxybutyl (R)-3-hydroxybutyrate) ingestion (KE) during a 3-week endurance training overload period markedly blunted the development of over-reaching symptoms (Poffé et al., 2019). This included suppression of overload training-induced bradycardia both at rest and during exercise, as well as a prevention of energy deficiency by increasing energy intake. KE did not impact sustainable training load during the first 2 weeks, but resulted in a 15% higher load in the final week of the intervention. This study for the first time demonstrated the potential of exogenous ketosis to impact training adaptation.
Interestingly, studies in mice also showed that ketone bodies, in particular β-hydroxybutyrate (βHB) and acetoacetate (AcAc), enhanced lymphangiogenesis (García-Caballero et al., 2019). Furthermore, βHB, but not AcAc, also stimulated the synthesis of vascular endothelial growth factor (VEGF) in mouse brain microvascular endothelial cells (Isales et al., 1999) and in the aorta of diabetic rats (Wu et al., 2020). This is of primary importance as exercise-induced angiogenesis is mainly controlled by VEGF (Gorski & De Bock, 2019). Addition of βHB in the micromolar range (1-10 mM) has also been shown to increase endothelial nitric oxide synthase (eNOS), a downstream factor of VEGF, in human endothelial cells during low-glucose-induced endothelial cell damage (Soejima et al., 2018). Furthermore, treatment of cardiac endothelial cells with both the ketone bodies βHB and AcAc directly promoted cell proliferation and sprouting capacity in vitro (Weis et al., 2022). Given these pro-angiogenic effects of ketone bodies seen in diverse tissues, it is reasonable to postulate that ketone bodies may also stimulate de novo capillarization in skeletal muscle. Interestingly, a recent study showed that acute KE substantially elevated serum erythropoietin (EPO) levels post-exercise (Evans et al., 2023). Although the role of EPO in promoting angiogenesis in healthy human skeletal muscle has been questioned (Lundby et al., 2008), EPO clearly stimulates angiogenesis under pathological conditions in humans (Nachbauer et al., 2012) as well as upon injury in rodents (Lamon & Russell, 2013). As such, this opens up another interesting mechanism by which KE could increase muscular angiogenesis in the catabolic context of overload training.
Therefore, the primary aim of this study was to characterize the mechanistic effect of KE on muscular angiogenesis. For this purpose we used the muscle biopsy samples obtained in our recent study published in this journal showing a wide spectrum of beneficial physiological effects due to intermittent KE during endurance training overload (Poffé et al., 2019). Furthermore, we explored the potential role of VEGF and EPO in modulating KE-induced capillary adaptation during training.

Subject recruitment and ethical approval
Twenty healthy, recreationally active male subjects participated in this study. Two subjects withdrew within the first week of the study, one due to adverse reactions to the protein-carbohydrate drinks prescribed by the study protocol, and one for reasons unrelated to the experimental procedures. Therefore, the data analyses presented in this report included 18 subjects (mean ± SD: age 21.3 ± 2.6 years, height 1.81 ± 0.04 m, body mass 73.7 ± 8.5 kg, maximal oxygen uptake (V O 2 max ) 55.7 ± 5.7 ml kg −1 min −1 ) who completed the full study protocol. During the entire study period, subjects were asked to abstain from any strenuous exercise other than prescribed by the study protocol. All subjects gave their written informed consent before inclusion in the study, which was approved by the Ethics Committee Research UZ/KU Leuven (registration number: B322201733747). The study conformed to the standards set by the Declaration of Helsinki, except for registration in a database.

Study design
The venous blood samples and muscle biopsy samples used for this study were obtained in the context of a larger project involving a randomized, double-blind J Physiol 601.12 experimental design of which the details have been reported elsewhere (Poffé et al., 2019). Briefly, subjects were pair-matched forV O 2 max , exercise performance, training history, body mass and height. Thereafter, subjects were enrolled in a fully controlled 3-week cycling training programme aimed to cause a state of over-reaching. During the training period, subjects received either a ketone ester drink (KE, n = 9) or a corresponding control drink (CON, n = 9). Muscle biopsies and venous blood samples were obtained on the day before (Pre) and after (Post) the overload training period.

Supervised training intervention
The training programme consisted of two training sessions per day (6 days/week) and included a combination of high-intensity interval training (HIIT), intermittent endurance training (IMT) and constant-load endurance training (ET) sessions (for detailed overview see Poffé et al., 2019). The training load was gradually increased over the 3-week period (week 1: ∼4600 kJ, week 2: ∼6400 kJ, week 3: ∼9600 kJ). Total training load was similar between KE and CON during week 1 (KE: 4665 ± 449 vs. CON: 4551 ± 436 kJ) and week 2 (KE: 6480 ± 684 vs. CON: 6355 ± 993, both P = 0.915), but in the final training week training load was 15% higher in KE (10,266 ± 962 kJ) than in CON (8962 ± 1939 kJ, P = 0.003). Each training session was performed in the laboratory on calibrated cycling ergometers (Avantronic Cyclus II, Leipzig, Germany and Tacx Neo Smart, Wassenaar, The Netherlands) and under the careful supervision of at least one of the investigators.

Post-exercise nutritional supplementation
Thirty minutes following each exercise session, subjects received a 500 ml high-dose protein-carbohydrate drink delivering 61 g carbohydrates and 31 g protein (6d Sports Nutrition, Oudenaarde, Belgium). In addition, subjects from KE received 25 g of ketone ester (>96% (R)-3-hydroxybutyl (R)-3-hydroxybutyrate; TdeltaS Ltd, Thame, UK) immediately following each training session and 30 min pre-sleep, whilst CON ingested an isocaloric control drink (CON) containing 16.4 g pure medium-chain triglycerides (Now Foods, Bloomingdale, IL, USA). To achieve a similar taste and appearance, bitter sucrose octaacetate (1 mM; Sigma-Aldrich, Bornem, Belgium) was added to CON, whilst a red colorant (AVEVE Bloem, Merksem, Belgium) was added to both the drinks. Immediately following ingestion of KE or CON, subjects received 50 ml of Diet Coke for mouth rinsing. The KE dosing strategy was successful to increase blood βHB levels up to ∼3 mM within 30 min post-exercise, whereas blood βHB levels remained low in CON (<0.5 mM). At Pre and Post, blood ketone levels were below 0.5 mM in both groups (Poffé et al., 2019).

Muscle biopsy sampling
On the day before (Pre) and after (Post) the training period, and 10−13 h after the last training session, a percutaneous needle biopsy (100-200 mg) from the midportion of m. vastus lateralis was obtained under local anaesthesia (2% xylocaine without adrenaline, 1 ml subcutaneously). Biopsies during Pre and Post were taken in the rested state from the left and right leg, respectively. After removal of visible blood, muscle samples were separated into two parts. One part was immediately frozen in liquid nitrogen and stored at −80°C for further biochemical analyses. Remaining muscle was mounted in embedding medium (Tissue-Tek OCT), frozen in precooled isopentane, and kept at −80°C until later histochemical analyses. To minimize potential diet-induced variations in muscle metabolism, subjects received a standardized carbohydrate-rich dinner (∼5400 kJ; 69% carbohydrate, 16% fat, 15% protein) and breakfast (∼2700 kJ; 71% carbohydrate, 15% fat, 14% protein) on the evening before and 1.5 h before the biopsy samplings, respectively.
Capillarization. Uncoated glass slides with serial cryosections (7 μm) from Pre and Post biopsies were air dried for 20 min at room temperature, followed by prehybridization in 1% BSA in PBS for 60 min. Thereafter, slides were incubated for 2 h at room temperature with a primary antibody against CD31 (1:500 in 0.5% BSA in PBS; DakoCytomation, AB_2114471). Following three 5 min washes (0.5% BSA in PBS), a biotinylated rabbit anti-mouse IgG (H&L) antibody (1:500; DakoCytomation, AB_2687571) was added for 1 h. Slides were then washed for 3 × 5 min and incubated for 1 h at 37°C with primary antibodies against myosin heavy chain I and IIa (BA-F8 (AB_10572253) and SC-71 (AB_2147165), respectively, dilution 1:200, Developmental Studies Hybridoma Bank). After washing (2 × 5 min), slides were incubated with the appropriate secondary antibodies (type I: goat anti-mouse IgG2b Alexa Fluor 488 (AB_2338856), type IIa: goat anti-mouse IgG1 Cy5 (AB_2338715), both 1:500, Jackson Immuno-Research Laboratories Inc.) during 1 h at 37°C. After three additional 5 min washes, streptavidin was applied for 30 min (1:100 in PBS, PerkinElmer, Waltham, MA), sections were washed again (3 × 5 min) and incubated with cyanine (1:100 in PBS, PerkinElmer) for 10 min. After a final washing step, sections were covered using Fluorescent Mounting Medium (DakoCytomation). Cryosections were captured at ×20 by fluorescence microscopy and subsequent images were analysed with ImageJ by an investigator blinded to subject coding. The number of capillaries was counted for each interior muscle fibre by a blinded investigator and expressed as capillary contacts. A random part of the capillarization images (∼30%) was also quantified by a second assessor. Bland-Altman analyses indicated a high degree of agreement between both assessors: capillary-to-fibre perimeter exchange index (CFPE, capillaries per 1000 μm perimeter; −0.239 ± 2.245%; 95% limits of agreement: −4.639 to +4.161%) and capillary contacts (−0.576 ± 2.217%; 95% limits of agreement: −4.921 to +3.768%). In addition, the CFPE was calculated in agreement with Hepple (1997). Briefly, the CFPE is calculated by dividing the capillary-to-fibre ratio, corrected for the number of fibres sharing each capillary (e.g. the capillary sharing factor), by the fibre perimeter. On average, 229 ± 74 (mean ± SD) fibres were analysed per cryosection. RNA extraction, reverse transcription and real-time quantitative PCR analysis. Total RNA was extracted from 15−20 mg frozen muscle tissue using TRIzol (Thermo Fisher Scientific, Vilvoorde, Belgium). Following isolation, the quality and quantity of the RNA were determined using a SimpliNano spectrophotometer (GE Healthcare, Chicago, IL, USA). cDNA was synthesized from 0.9 μg RNA using the cDNA Reverse Transcription kit (Qiagen, Hilden, Germany) according to the manufacturer's recommendations, which included a genomic DNA wipeout step. Real-time quantitative PCR was performed using the GoTaq qPCR Master Mix (Promega, Madison, WI, USA) on a QuantStudio 3 thermocycler (Thermo Fisher Scientific, Waltham, MA, USA). Real-time PCR primers were designed for human VEGFa, eNOS and potential housekeeping genes cyclophilin A (CycloA), and ribosomal protein L19 (RPL19). Primer sequences were as follows: CycloA forward, CTT CAT CCT AAA GCA TAC GGG TC; CycloA reverse, TGC CAT CCA ACC ACT CAG TCT; RPL19 forward, CGC TGT GGC AAG GTC; RPL19 reverse, GGA ATG GAC CGT CAC AGG C; VEGFa forward, TTT CTG TCT TGG GTG CAT TGG; VEGFa reverse, ACC ACT TCG TGA TTC TGC CCT; eNOS forward, CAG TTA CCA GCT AGC CAA AGT; eNOS reverse, CTC ATT CTC CAG GTG CTT CAT. The cycle threshold value for CycloA was affected by neither training (P = 0.677) nor KE (P = 0.336), and was relatively stable (CV: 4.43%). In contrast, the cycle threshold value for RPL19 tended to decrease with training (P = 0.085). Therefore, we decided to use only CycloA as housekeeping gene. Each sample was run in duplicate and gene expression was calculated using the delta threshold cycle method with CycloA as the housekeeping gene.
Blood sampling and analysis. At Pre and Post, venous blood samples were obtained in the fasted state from an antecubital vein (Venoject, Terumo, Tokyo, Japan) and collected into vacuum tubes containing Silica Clot Activator (Becton Dickinson (BD) Vacutainer, Eysins, Switzerland). Tubes were centrifuged (1500 rpm for 10 min at 4°C) and the supernatant was stored at −20°C until later analysis. A commercially available enzyme-linked immunosorbent assay was performed to determine serum EPO levels (DY286-05, R&D Systems, Minneapolis, MN, USA).
Statistics. Statistical analyses were performed in GraphPad Prism version 9.3.1 (GraphPad Software, La Jolla, CA, USA). A two-way repeated measures analysis of variance (ANOVA) (group × time) was performed to assess differences in mean values between the experimental groups and over time, followed by a Bonferroni post hoc test if a significant interaction was detected. Reported P-values refer to the output of the post hoc analyses in case of a significant interaction effect, whilst otherwise P-values for main effects are reported. Data were checked for equal variance (Brown-Forsythe P > 0.05) and normality of residuals (Kolmogorov-Smirnov P > 0.05). For the data that did not comply with these assumptions (percentage type IIa and IIx fibres, cross-sectional area type I and IIa fibres, VEGF mRNA expression, phospho/total AMPKα, PGC1α and ERRα) we applied a log transformation. After this transformation, all parameters except VEGF mRNA expression complied with the assumptions and were analysed by a two-way repeated-measures ANOVA. VEGF mRNA expression data were transformed by taking the square root, which resulted in their complying with the assumptions, and were subsequently analysed by a two-way repeated-measures ANOVA. Data used for correlation analyses were normally distributed (Kolmogorov-Smirnov P > 0.05), with Pearson's correlation coefficient calculated using percentage change scores. A probability level of P < 0.05 was set as statistically significant. Data are expressed as means ± SD. Effect sizes (ES) are shown as partial eta-square (η p 2 ) for main and interaction effects, and in cases of post hoc pair-wise comparisons as Hedge's g using untransformed data (Lakens, 2013). The primary outcome of this study is muscle capillarization (capillary contacts and CFPE-index), whereas skeletal muscle VEGF and serum EPO were defined as secondary outcomes. All other measurements are tertiary outcomes.

Muscle capillarization, fibre type composition and cross-sectional area
A time × KE interaction effect was detected for both the number of capillary contacts (CC; P = 0.048, η p 2 = 0.22) and the capillary-to-fibre perimeter exchange index (CFPE; P = 0.031, η p 2 = 0.26, Fig. 1A-C). The training period altered neither CC (P > 0.999) nor CFPE (P = 0.333) in CON. In contrast, in all subjects on KE the training did increase both capillarization indices. Compared with baseline, at the end of the training period CC and CFPE on average were increased by 44% (P = 0.005, g = 1.40) and 42% (P < 0.001, g = 1.78), respectively. This resulted in higher CFPE (P = 0.038) and a tendency for more CC (P = 0.092) in KE compared to CON at Post. The increase in both CC and CFPE from Pre to Post in KE was evident in both type I (P = 0.005 for both CC and CFPE) and type II (CC: P = 0.002; CFPE: P < 0.001) muscle fibres. Conversely, in CON CC and CFPE were unaffected in both type I (CC: P > 0.999; CFPE: P = 0.638) and type II (CC: P = 0.870; CFPE: P = 0.232) fibres. Neither an interaction nor a main effect of KE was observed for fibre type distribution, mean fibre cross-sectional area (CSA) or fibre type specific CSA in m. vastus lateralis (Table 1). A main effect of time was only observed for the percentage of type IIx fibres (P = 0.011), indicating that the percentage of type IIx fibres decreased in response to the training.

Muscle biochemistry
To determine the molecular mechanisms promoting the angiogenic effect of KE, we first assessed the proangiogenic protein VEGF. A time × KE interaction effect was observed for VEGF both at the mRNA (P = 0.038, η p 2 = 0.24, Fig. 2A) and the protein (P = 0.032, η p 2 = 0.26, Fig. 2B) level. In KE, the training increased muscle VEGF protein content by 33% (P = 0.028, g = 0.61), while mRNA expression increased 2-fold (P = 0.008, g = 0.97). In contrast, VEGF mRNA and protein expression were unchanged in CON (both P > 0.999). For eNOS, a downstream target of VEGF, a time × KE interaction effect was observed at the mRNA level (P = 0.006, η p 2 = 0.38, Fig. 2C). eNOS mRNA expression increased twofold in response to training in KE (P < 0.001, g = 1.49), but was unaffected in CON (P > 0.999). Conversely, only a main effect of time (P = 0.001, η p 2 = 0.49, Fig. 2D) was observed for eNOS protein expression indicating that eNOS increased by 125% (g = 1.03) at the protein level in both KE and CON in response to training. To further characterize the effect of KE on exercise-induced angiogenesis, we assessed upstream transcription factors of VEGF implicated in both baseline (AMPK) and exercise-induced (PGC1α, HIF1α and ERRα) Figure 1. Ketone ester intake stimulates muscle capillarization in response to overload training A, representative immunofluorescence images of m. vastus lateralis before (Pre) and after (Post) training in healthy, male subjects receiving ketone ester (KE, n = 9) or control (CON, n = 9). CD31, red; type I, green; type IIa, blue; type IIx, black. Scale bar: 100 μm. B and C, dot plot together with mean (circle) ± SD (whiskers) representing capillary-to-fibre perimeter exchange index (CFPE-index) (B) and capillary contacts (C) in KE (•, n = 9) compared to CON (•, n = 9) following training. Images were captured at ×20 magnification (Nikon E1000, Nikon, Boerhavedorp, Germany). No adjustments were applied to the channel capturing the capillaries, whereas the brightness of the channels capturing the type I and type II fibres was similarly adjusted for the Pre and Post condition of each individual by an investigator who was blinded to the experimental conditions. Two-way repeated-measures ANOVA with a Bonferonni post hoc test. * P < 0.05 KE vs. CON  Values are means ± SD and represent fibre type distribution and fibre cross-sectional area (CSA) of m. vastus lateralis before (Pre) and after (Post) overload training in male subjects receiving ketone ester (KE, n = 9) or control (CON, n = 9). # P < 0.05 vs. Pre for indicated group.
angiogenesis at the protein level. A main effect of time was observed for both phospho-AMPKα Thr172 (P = 0.017, η p 2 = 0.31, Fig. 3A) and total AMPKα (P = 0.021, η p 2 = 0.29, Fig. 3B), indicating that both phospho-and total AMPKα increased by ∼40% in response to training in both KE and CON. Except for a small increase with training for ERRα in both groups (P = 0.030, η p 2 = 0.26), none of the other proteins were affected by either the training or KE (all P > 0.05, Fig. 3C-F).

Serum erythropoietin levels
Next we determined whether KE also impacted serum EPO, which is a potent angiogenic factor. A time × KE interaction effect (P = 0.043, η p 2 = 0.23, Fig. 4A) was detected for serum EPO. Serum EPO concentration was stable in response to training in CON (P > 0.999), but increased by ∼25% in KE (P = 0.027, g = 0.70). Serum EPO changes were closely correlated with changes in muscle VEGF protein expression (r = 0.749, P < 0.001, Fig. 4B).

Discussion
An increase in skeletal muscle capillarization is one of the central adaptations to exercise training, and precedes other adaptive responses in skeletal muscle such as an increase in oxidative capacity (Škorjanc et al., 1998). However, such beneficial adaptations may be inhibited by excess training load. Interestingly, we recently observed that ketone ester ingestion (KE) counteracts physiological dysregulations induced by overload training. With this background, here we investigated whether KE could stimulate muscular angiogenesis during a 3-week overload training period. Skeletal muscle capillarization did not increase in response to the overload training period in the control group (CON), but increased by ∼40% in the KE condition. In addition, this increase was accompanied by higher circulating erythropoietin concentration as well as stimulation of multiple pro-angiogenic factors such as muscle VEGF and eNOS.
Skeletal muscle capillarization indices (e.g. capillary contacts and capillary-to-fibre perimeter exchange index) did not increase in response to the 3-week endurance overload training period in CON. Nonetheless, a recent meta-analysis (Liu et al., 2022) indicated that skeletal muscle capillarization already increases within only 2 weeks of endurance exercise training in untrained subjects, with an average increase in capillary to fibre ratio of ∼20% within 2−4 weeks. This thus indicates that overload training impaired the normal angiogenic response to endurance exercise. As such, our data are in line with earlier reports showing that over-reaching impairs exercise-induced improvements in muscle oxidative capacity (Bellinger et al., 2019) and endurance exercise performance (Aubry et al., 2014). This observation also provides further support for a previous study showing that (too) intense exercise delivers a weaker stimulus for endothelial cell proliferation and VEGF secretion compared to moderate intensity exercise (Hoier et al., 2013).
In contrast to CON, all subjects in KE exhibited enhanced skeletal muscle capillarization in response to the overload training period. This adaptation occurred independent of alterations in either muscle fibre morphology or fibre type distribution. This either indicates that KE counteracted the development of over-reaching and thereby enabled the normal angiogenic response to exercise to occur, or that KE directly stimulated angiogenesis. In favour of the first hypothesis, we previously reported KE to blunt the development of numerous over-reaching symptoms in the context of the current study (Poffé et al., 2019). Nonetheless, previous data obtained in rodents have shown that ketone bodies can directly stimulate lymphangiogenesis (García-Caballero et al., 2019) and can upregulate multiple pro-angiogenic factors such as VEGF (Isales et al., 1999;Wu et al., 2020). In line with the latter, we observed that KE increased skeletal muscle VEGF expression both at the protein and at the mRNA level. This is a key finding because myocyte-expressed VEGF is a prerequisite for training-induced muscular angiogenesis (Delaver et al., 2014;Olfert et al., 2010). In addition, exercise-induced angiogenesis is mainly controlled by VEGF (Gorski & De Bock, 2019). This suggests that the exercise-induced angiogenic response in KE was mediated by VEGF. Furthermore, the absence of an increase in skeletal muscle VEGF in CON supports the prevailing opinion that excessive exercise intensity blunts exercise-induced VEGF production, which in turn may impair capillary growth (Gliemann et al., 2015;Hoier et al., 2013).
Interestingly, the increase in VEGF mRNA expression in KE occurred more than 10 h following the last training session, while previous studies indicated that the exercise-induced stimulation of VEGF mRNA fades within 6 h (Breen et al., 1996;Hoier & Hellsten, 2014;Jensen et al., 2004). The elevated VEGF mRNA and protein expressions in KE also occurred 9−12 h after the last ketone ester dose, i.e. at identical low (<0.5 mM) blood ketone body concentrations between KE and CON. The effect of intermittent KE to stimulate VEGF thus most likely reflects a short-term adaptation which occurred independent of acute regulation by KE, and thus likely represents a mechanism distinct from the role of ketone bodies as an energy substrate. Nonetheless, it should be acknowledged that previous studies only measured VEGF mRNA either within the first 6 h post-exercise or 24 h after exercise, and as such we could not exclude the possibility that exercise per se increased VEGF mRNA ∼10 h after exercise. This is in contrast with the ability of ketone bodies to stimulate lymph vessel growth, which is dependent on the ketolytic enzyme OXCT1 (García-Caballero et al., 2019). In support of a signalling role, the KE-induced increase in skeletal muscle capillarization in our study was similar between type I and type II fibres, despite type I fibres having a higher capacity for sarcolemmal ketone body transport (Bonen, 2001) and oxidation (Winder et al., 1974). Nonetheless, given the limited data available on the time course of VEGF following exercise, and the fact that muscle biopsies were collected 10−13 h following the last training session, it cannot be excluded that the KE-induced increase in VEGF also includes a residual effect from this last training session. However, training involves a sequence of acute response to individual exercise sessions, and it is the repetition of these responses that eventually causes the training adaptation (Booth & Thomason, 1991). Therefore, even if the increase in VEGF represents an acute response, it is still crucial for the adaptive response to training. To further characterize the effect of KE on exercise-induced angiogenesis, we assessed upstream transcription factors of VEGF implicated in both baseline (AMPK) and exercise-induced (PGC1α, HIF1α and ERRα) angiogenesis (Gorski & De Bock, 2019). However, none of these factors differed between KE and CON. Nonetheless, this does not per se indicate that these metabolic regulators do not mediate the KE-induced activation of VEGF as they for instance might have been altered only during the first few hours in recovery, and as such were not detected given that muscle biopsies were taken >10 h post-exercise. In this regard, a previous study showed that endurance exercise-induced phosphorylation of AMPK fades within 3 h after exercise (Bartlett et al., 2012).
Interestingly, we observed that KE increased serum EPO levels in response to the training period. Such an increase in serum EPO has recently been observed under acute ketosis post-exercise (Evans et al., 2023). However, our data indicate that elevation of circulating EPO concentration can persist for many hours after blood ketone levels have returned to baseline. Although EPO is generally known for its haematopoietic effects, it can also exert pro-angiogenic effects on skeletal muscle under pathological conditions (Lamon & Russell, 2013;Nachbauer et al., 2012). Experiments in rodents as well as observations in endothelial cells from bovine aorta indicate that EPO can directly upregulate VEGF (Kimáková et al., 2017;Nakano et al., 2007), while the other way around, VEGF is also a potent EPO inducer (Greenwald et al., 2019). Taken together with the strong correlation (r = 0.75) between the training-induced increase in serum EPO and muscle VEGF protein expression seen in our study, it is tempting to speculate that KE may stimulate muscular angiogenesis via EPO-induced activation of VEGF. Nonetheless, other mechanisms such as an upregulation of VEGF via ketone oxidation-induced elevations of intramyocellular NAD + (Das et al., 2018;Elamin et al., 2017) or intramyocellular calcium concentrations (Isales et al., 1999) may also be involved.
The most highlighted action of EPO is stimulation of erythropoiesis, leading to elevation of total circulating haemoglobin mass, which is a primary determinant ofV O 2 max (Schmidt & Prommer, 2010). Because we measured neither total haemoglobin mass nor the changes ofV O 2 max in the current protocol, we cannot evaluate whether the small increment in EPO measured at the end of the training (∼25%) reflects such an ergogenic mechanism. However, a recent study demonstrated that three microdoses of human recombinant erythropoietin during 4 weeks was sufficient to raise total haemoglobin mass, increaseV O 2 max , as well as improve cycling time-trial performance in trained volunteers (Andersen et al., 2023). Furthermore, beyond stimulating VEGF and haematopoiesis, EPO has also been shown to induce multiple other responses in a variety of tissues that are relevant to post-exercise recovery (Suresh et al., 2020). In mice, EPO has been shown to improve recovery from cardiotoxin-induced skeletal muscle injury by increasing the number of satellite cells (Jia et al., 2012). EPO also has been shown to provide protective effects on the heart and the brain during ischaemic stress. Furthermore, EPO decreased proinflammatory cytokines and macrophage infiltration in white adipose tissue, and inhibited pituitary adrenocorticotropic hormone (ACTH) secretion (Suresh et al., 2020). Taken together, literature findings and the current observations suggest that EPO might be implicated in the potential of KE to enhance recovery following strenuous exercise via a multiplicity of actions in different tissues. Nonetheless, further mechanistic studies Figure 4. Ketone ester intake increases serum erythropoietin levels A, dot plot together with mean (circle) ± SD (whiskers) for serum EPO levels before (Pre) and after (Post) overload training. B, correlation between the percentage change in serum EPO and skeletal muscle VEGF protein expression. During the training period male subjects ingested either a control drink (•, n = 9) or ketone ester supplements (•, n = 9). Two-way repeated-measures ANOVA with a Bonferonni post hoc test. #P < 0.05 vs. Pre for indicated group. EPO, erythropoietin; VEGF, vascular endothelial growth factor. J Physiol 601.12 are required to identify the biological relevance of the observed increase in circulating EPO upon KE.
We previously reported that, compared with CON, KE in the conditions of the current study increased dietary energy intake by ∼20%, as well as allowing for 15% higher training load to be tolerated during the final week of the training intervention (Poffé et al., 2019). Hence it may be argued that the angiogenic effect of KE during the training overload period was at least partly, if not largely, due to differential energy balance and training load between groups. To our knowledge there are no data available to indicate whether or not slight short-term modifications in food intake, energy balance or training load, as occurred in the current study, can impact the angiogenic response to exercise in young, healthy subjects. In fact, an increase in training load during strenuous training probably would rather inhibit than stimulate beneficial muscular adaptations (Bellinger et al., 2019). In fact, training load and dietary energy intake were identical during the initial 2 weeks of training, and from a physiological perspective it is impossible to explain an ∼40% increment in capillarization by the small differences in energy intake and training load that occurred during the third week of training only. But it remains to be established whether the angiogenic response to KE during training seen here would also be present to a similar degree if energy intake and training load were perfectly matched between the experimental conditions, and whether it is specific for the current conditions of training overload, or would also emerge during training that is balanced by adequate recovery.
In conclusion, we for the first time demonstrate that oral ketone ester administration markedly stimulates muscular angiogenesis during endurance training overload. This most likely resulted from a direct stimulation of angiogenesis via upregulation of erythropoietin secretion and muscular VEGF activity, together with an inhibition of overload training-induced impairments of the normal angiogenic response to exercise. These observations provide novel strong evidence to support the potential of exogenous ketosis to facilitate physiological and functional recovery from strenuous endurance exercise. Further research should identify whether the angiogenic response to KE represents a novel direct signalling effect of ketone bodies, or was due to the specific conditions of the current study involving endurance training overload in conjunction with ketone ester intake.