Increased capillary tortuosity and pericapillary basement membrane thinning in skeletal muscle of mice undergoing running wheel training

Two tissues/organ systems in particular perceive the systemic impact provoked by regular physical activity (such as running or cycling training) that might significantly improve physical fitness and thus positively influence the quality of life, including extension of lifetime. First, endurance exercise triggers adaptive changes in the structure and function of the skeletal muscle fibers, i.e. by mitochondrial biogenesis (Holloszy, 1975; Hood et al., 2006) and by induction of fiber-type shifting without hypertrophy (Pette and Staron, 1997). Most molecular mechanisms that have been identified to date that contribute to physiological responses to endurance exercise are attributed to skeletal muscle fibers (Hoppeler et al., 2011). Second, a significant proportion of the positive effects evoked by regular physical activity are related to the cardiovascular system (Hellsten and Nyberg, 2016). In particular, the heart and larger-sized blood vessels may functionally adapt to endurance exercise and thereby contribute to health improvements (Hellsten and Nyberg, 2016; Laughlin, 2016). However, it also appears likely that the capillaries (as the unit of the vascular system with the smallest diameter) may undergo microvascular remodeling in response to a continuous training stimulus. Consequently, the microcirculation may supply peripheral tissues with oxygen and nutrients and may remove carbon dioxide and catabolic products, respectively, in a more efficient way.

The most prominent example for such an endurance exercise-induced microvascular remodeling analyzed so far is the increase in the numerical density of the capillaries, which is a process being designated angiogenesis (Hudlicka, 1998; Olfert et al., 2016). Other adaptive changes of the capillary system structure in skeletal muscles in response to endurance exercise (or chronic electrical stimulation, an animal model which resembles endurance exercise) have hitherto only been described sporadically, such as the reduction of the basement membrane (BM) thickness around capillaries (CBMT) (Baum and Bigler, 2016; Williamson et al., 1996) as well as transient short-term endothelial cell (EC) thinning (Peeze Binkhorst et al., 1989) and late-stage EC swelling (Egginton and Hudlická, 1999). A systematic synopsis of the structural adjustments of skeletal muscle capillaries to endurance exercise still needs to be performed.

In order to understand the dynamics of microvascular remodeling in response to endurance exercise, it is helpful to regard the regulation of the capillary system phenotype to fulfill its carrier function as a negative feedback control circuit. According to this cybernetic concept, the system is represented by the capillary function as the exchanger, the sensor is denoted by several molecular systems and the controller is given by the capillary phenotype. To understand this concept, it is helpful to consider some molecular players that have already been identified to operate in such negative feedback control circuits. If the microcirculation is not sufficiently structured to fulfill the metabolic demands imposed on the musculature (e.g. during/after endurance exercise), the oxygen partial pressure reduces and/or the concentrations of energy substrates or carriers become too low. The dysfunction is sensed by ECs and/or the muscle fibers [e.g. by the prolyl-4-hydroxylase domain proteins/hypoxia-inducible factor oxygen sensing system; 5′-AMP-activated protein kinase (AMPK), sirtuins, peroxisome proliferator-activated receptors, soluble guanylate cyclase (for overview, see Freyssenet, 2007; Hoppeler et al., 2011)]. This information is subsequently converted into an altered gene expression profile [control variable; e.g. by changing the activity levels of the molecular AMPK/PGC-1α/VEGF axis (Leick et al., 2009)], which then alters the EC phenotype (downstream output). If the metabolic homeostasis is re-established, the adjustment of the capillary phenotype is not continued or might be reversed. This model demands that the structural phenotype of the capillary system in skeletal muscle is tightly regulated and its plasticity is relevant for the correct function of the muscular tissue according to the basic requirement of biology that ‘function follows form’. We therefore consider it crucial to exactly understand the mechanisms of how the phenotype of the capillary system is formed at different stages of the adaptive process to endurance exercise activity.

Recently, we have characterized the ultrastructure of capillaries in skeletal muscle of humans before and after an 8 week period of endurance exercise (Baum et al., 2015). The intense ergometer training of the study participants was accompanied by angiogenesis in the vastus lateralis muscle biopsies, which was statistically related to increased microcirculatory pericyte (PC) coverage and thinning of the CBMT (Baum et al., 2015). We furthermore observed a significant volume expansion of the capillary ECs in the muscle biopsies, which was not related to the onset of angiogenesis (Baum et al., 2015). However, these findings represent only end-stage observations and do not provide information about early adaptive responses of the capillaries to the training stimulus.

In continuation of this previous investigation performed on human skeletal muscle biopsies (Baum et al., 2015), we have assessed whether the structure of skeletal muscle capillaries is likewise changed in mice exposed to endurance training. In particular, we hypothesized that the structural changes of the capillary organization in skeletal muscle of mice are: (1) similar to those in humans after a long period of endurance training; and (2) already manifested in the early stages of the training. To verify these hypotheses, we subjected C57BL/6 mice to voluntary running wheel training for 1 week (1 wk-t) or 6 weeks (6 wks-t) and quantitatively described the capillarity in the plantaris (PLNT) muscle as well as the compartmental and sub-compartmental organization of capillaries in comparison to that of untrained control mice.

Eighteen male C57BL/6 mouse (Mus musculus Linnaeus) littermates aged 12 weeks (purchased from Charles River, Sulzfeld, Germany) were randomly allocated to one of three groups: (1) sedentary control mice; (2) mice trained for 1 week (1 wk-t); and (3) mice trained for 6 weeks (6 wks-t).

All mice were maintained in a conventional animal facility in Bern, Switzerland, with a fixed 12 h:12 h light:dark cycle on a commercial pelleted chow diet with free access to tap water. At sacrifice, mice were anesthetized with a ketamine/xylazine (100 mg kg⁻¹/5 mg kg⁻¹) cocktail via intraperitoneal injection. The euthanasia of all mice was carried out within two days. The experiments were performed in accordance with the approvals published by the Cantonal Committee on Animal Welfare [Amt für Landwirtschaft und Natur des Kantons Bern (27/12)] and the University of Bern.

All mice were housed individually in cages each equipped with an 18 cm-diameter impeller purchased from a local pet shop (Fressnapf, Dietikon, Switzerland) and a magnetic revolution counter (in-house manufacturing with components obtained from Conrad, Dietikon, Switzerland). The revolution counters were read and reset to zero daily at 08:30 h and 17:00 h. To calculate the running distances (in m), the number of rotations was multiplied by 2×π×0.09 (the latter value is the radius of the impeller in m).

PLNT muscle samples were chemically fixed in a 6.25% (v/v) glutaraldehyde solution buffered with 0.1 mol l⁻¹ sodium cacodylate–HCl (pH 7.4) and stored at 4°C until analysis.

The chemically fixed PLNT muscle samples were divided into 4–5 pieces, each with a volume of ∼0.5 mm³, after which they were post-fixed in 1% (w/v) OsO₄, stained en bloc and embedded in Epon 812 (Fluka, Buchs, Switzerland). One micrometer ‘semi-thin’ sections were cut using a diamond knife and stained with 0.5% (w/v) Toluidine Blue dissolved in 1% (w/v) sodium tetraborate for 15 s.

For the morphometric evaluation of capillarity, transverse sections through the muscle (size of ∼1 mm²) were cut from two randomly selected Epon blocks from each PLNT muscle. A systematic sampling strategy was implemented to acquire six light micrographs of each section at a magnification of ×400 in a Leica DMR light microscope (Leica Microsystems, Heerbrugg, Switzerland). The light microscope was equipped with a programmable motor-driven x/y sampling stage allowing defined stepwise movements to sample image fields in a systematic uniform random way. This equipment ensured that the micrographs, which we recorded for the morphometric analysis, embody non-overlapping areas representative of the entire muscle cross-section. Subsequently, the Epon blocks were turned 90 deg to prepare longitudinal sections, which were always large enough to gain six light micrographs taken by the same protocol mentioned above.

On the light micrographs of the transverse PLNT muscle sections, the number of capillary profiles and that of muscle fiber profiles were counted, taking into account the forbidden line rule (Weibel, 1979). The mean cross-sectional fiber area (MCSFA) was estimated by relating the area on the micrographs covered by skeletal muscle fiber profiles (which was assessed by point counting on a 10×10 point grid with each point representing an area of 0.365 µm²) to the number of muscle fiber profiles. The capillary-to-fiber (C/F) ratio was computed as the number of capillary profiles divided by the number of skeletal muscle fibers, whereas the capillary (profile) density on transverse sections N_A(c,f)=Q_A(0) was calculated as the number of capillary profiles divided by the section area covered by skeletal muscle fiber profiles.

The sarcomere length was determined on the longitudinal PLNT muscle sections. Therefore, a minimum 100 µm-long reference line was drawn digitally along a muscle fiber profile orthogonal to the sarcomeric striation and in parallel to the sarcolemma. Densitometry was performed along this reference line to visualize the sarcomeric striation. The length of the reference line was related to the number of sarcomeric units in order to obtain the mean sarcomere length.

The dimensionless tortuosity factor on transverse sections c(K,0) was established following a morphometric protocol developed by Weibel, Mathieu-Costello and colleagues (Mathieu et al., 1983; Mathieu-Costello et al., 1989). This procedure takes into account the ‘Fisher axial distribution’ for directional anisotropy. Therefore, the ratio between the capillary density on the transverse sections Q_A(0) and the capillary density on the longitudinal sections Q_A(π/2) was calculated. Q_A(0)/Q_A(π/2) can be used to read out the concentration parameter K and the corresponding c(K,0) in Mathieu et al., 1983. For reasons of simplicity, this procedure might be abbreviated by the application of the polynomial function, which we have developed by making use of the data collection published by Mathieu et al., 1983: c(K,0)=−0.0011x⁵+0.0261x⁴−0.25x³+1.1709x²−2.7535x+3.7875, with x standing for Q_A(0)/Q_A(π/2). In our experience, this equation results in acceptable approximations for the tortuosity factor in ranges for Q_A(0)/Q_A(π/2) that exist in skeletal muscles of humans and rodents.

The capillary length density J_v was calculated by multiplication of the values for Q_A(0) and c(K,0).

Ultra-thin sections (50–60 nm in thickness) of the muscles were prepared with an Ultracut ultramicrotome (Reichert-Jung, Bensheim, Germany), floated on 200-mesh copper grids (Plano, Wetzlar, Germany) and contrasted with uranyl acetate and lead citrate. The inspection was carried out using a transmission electron microscope [(TEM) Morgagni M268; FEI, Brno, Czech Republic].

Twenty to twenty-five randomly depicted electron micrographs of capillary profiles per ultra-thin section were photographed in the TEM at a final magnification of ×7.800. Micrographs showing capillary profiles with a length-to-width ratio of the smallest and the longest diameter of more than 1.2 were considered to be too obliquely or longitudinally sectioned and were thus excluded from morphometric evaluation.

Tablet-based image analysis was performed for the capillary morphometry by two researchers (C.S. and A.R.). On 20 electron micrographs showing the capillaries, lines were drawn with a digital pen around the lumen (lumen/EC transition), along the abluminal EC surface (EC/BM transition), at the BM/endomysium transition and around the PC surface of the capillaries. By processing with ImageJ, the values for the profile areas (A_lumen, A_EC, A_PC, A_BM) and profile perimeters (P_{lumen/EC transition}, P_{EC/BM transition}, P_{BM/endomysium transition}) of the structures of interest were obtained and then the means of the two measurements were computed to gain structural indicators that describe quantitatively the capillary ultrastructure: the absolute cross-sectional area (A) of the capillary and each of its compartments; and the profile area density (A_A) of each compartment relative to the capillary profile area (A_lumen+A_EC+A_PC+A_BM). The absolute values for the radius of the lumen and the total capillary profile as well as the arithmetic thickness (T) of the endothelium and the BM were calculated as previously reported (Bigler et al., 2016).

The PC coverage of capillaries was estimated as the ratio of the length of the abluminal EC perimeter covered by a PC profile with the total abluminal EC perimeter, as previously reported (Egginton et al., 1996; Tilton et al., 1985). The intraluminal EC surface enlargement was calculated as the length of the luminal EC perimeter with EC protrusions divided by luminal EC perimeter without protrusions minus 1. For additional characterization of intraluminal EC surface enlargement, we related the number of capillary profiles with intraluminal protrusions/filopodia longer than 5.2 μm (which corresponds approximately to the doubled mean inner diameter of the capillaries) to the number of total capillary profiles.

The junctional interaction between ECs and PCs was assessed in accordance with previous reports (Allsopp and Gamble, 1979; Bigler et al., 2016; Egginton et al., 1996). Therefore, semi-quantitative indicators were computed by relating the number of capillary profiles exhibiting the sub-compartmental junctions of interest [i.e. projections of the PCs (‘PC pegs’) invading the ECs (‘EC sockets’) as well as intracellular holes in PCs (‘PC sockets’) caused by invading EC projections (‘EC pegs’), PC curling or PC–PC contacts] to the total number of capillary profiles analyzed.

Numerical data are expressed as mean values together with the standard deviations. All morphometric data sets were tested by Kolmogorov–Smirnoff with Lillefors correction and Shapiro–Wilk for their normality of distribution prior to statistical analysis. Comparisons pertaining to the morphometric analyses between control mice and mice of the 1 wk-t and 6 wks-t groups were checked using one-way ANOVA followed by pairwise post hoc Tukey's multiple comparison test. If the third value of a structural indicator (6 wks-t) was inversing the trend of the second value (1 wk-t), we additionally tested for statistical significance by performing a pairwise two-tailed Student's t-test, as effects reversed by extended running wheel training (discontinuous sequence) are not picked up by ANOVA. Statistical significance was assumed for: ANOVA α=0.05; Tukey's multiple comparison test α=0.05, 0.01 and 0.001, respectively; and Student's t-test α=0.05.

The performance of the mice forming the 6 wks-t group increased significantly during the second and third weeks of the training period (Fig. 1): their initial mean daily running distance of 5.3± 0.9 km improved 45% in the second week (compared with the first week) to increase by an additional 40% in the third week (compared with the second week). The changes in the daily running distance measured in the following weeks (5% in the fourth week, 17% in the fifth week and −3% in the sixth week; always in respect to the mean running distance of the previous week) were not significant. In total, the daily running distance of the mice increased 139% between the first and the sixth week of training. Taken together, the running activity of the mice improved for three weeks to merge into equilibrium at a high level for the residual training period. Impressively, some mice were active on the running wheel for ∼18 km per day after 6 weeks of exercising. We also want to mention that the mice in the 1 wk-t group ran 5.1±0.8 km daily on the running wheel (data not shown), which corresponds to the first week’s running activity of the 6 wks-t group.

On transverse sections (Fig. 2A,C,E) of the PLNT muscle, capillaries were identified as small round-shaped profiles mostly with visible lumen surrounding the skeletal muscle fiber profiles. On longitudinal sections (Fig. 2B,D,F), capillaries were distinguishable as round-shaped or oblique profiles (either isolated or grouped) that were elongated to a variable extent. The elongated capillary profiles appeared to preponderate in the PLNT muscle of the mice from the control group (Fig. 2B), whereas the round-shaped capillary profiles were more frequently observed in the PLNT muscle of the exercising mice (Fig. 2D,F), especially in the 6 wks-t mice (Fig. 2F). Furthermore, a regular striation was noticed inside the skeletal muscle fibers on the sections of all mice, which was caused by their sarcomeric organization (Fig. 2F, inset).

The transverse and longitudinal semi-thin sections of the PLNT muscle were subjected to morphometry (Fig. 3). The C/F ratio (32%; P<0.05) and the capillary density N_A(c,f) (36%; P<0.05) were higher in the 1 wk-t group than the control group. In the 6 wks-t group, C/F ratio (34%; P<0.01) and N_A(c,f) (23%; P=0.07) were likewise higher than in the 1 wk-t group and, thus in total, 76% (C/F ratio; P<0.001) and 66% [N_A(c,f); P<0.001] higher than in the control group. The MCSFA and the sarcomere length varied only non-significantly between the three groups (MCSFA: control versus 1 wk-t, −6.3%; control versus 6 wks-t, +2.3%; 1 wk-t versus 6 wks-t, 9.2%; sarcomere length: control versus 1 wk-t, −1.6%; control versus 6 wks-t, −3.9%; 1 wk-t versus 6 wks-t, −2.4%). c(K,0) in the PLNT muscle differed only non-significantly (6.7%) between the control and the 1 wk-t groups and (9.0%) between the 1 wk-t and 6 wks-t groups. In total, c(K,0) was significantly 16.3% higher in the PLNT muscle of the 6 wks-t mice compared with the control group. J_v in the PLNT muscle was significantly 43% higher in the 1 wk-t group than the control group and significantly 34% higher in the 6 wks-t group than in the 1 wk-t group, resulting in a significant 92% difference between the 6 wks-t group and the control group.

Whereas the transversely sectioned capillary profiles from the PLNT muscle of mice from the three study groups (Fig. 4A–C) were subjected to a morphometric analysis for the quantitative assessment of their compartmental composition [lumen, EC, BM and PC], the longitudinally sectioned capillary profiles were studied only qualitatively. Strikingly, we occasionally found large series of transversely sectioned capillaries to be girded in sarcolemmal pits in close proximity to densely packed subsarcolemmal mitochondria (Fig. 4D). Furthermore, we used the transverse capillary sections to assess semi-quantitatively the appearance of sub-compartmental peg–socket junctions (PC pegs–EC socket; EC pegs–PC socket) in the PLNT muscle of the mice from the three study groups. As seen in the examples shown in Fig. 4E–I, peg–socket junctions represent projections or filopodia of cells (‘pegs’) that curl into itself or invade other cells at their abluminal surface visible as pale pockets and holes in their cytoplasm (‘sockets’).

For each PLNT muscle, electron micrographs of 20 randomly selected capillaries were subjected to morphometry. As shown in Table 1, the profile area size belonging to the capillary lumen A (lumen) was larger (30%, P=0.06, P=0.04 in Student's t-test) in the PLNT muscle of 1 wk-t mice than controls. The profile area density of the BM was lower (−19.8%; P<0.05) after 1 week of training and after 6 weeks of training (−20.7%; P<0.05). Computation of the values for profile area sizes and perimeters revealed that the radius of the capillary lumen tended to be higher (17.6%; P=0.09, P=0.03 in Student's t-test) in the 1 wk-t group than the control group, whereas the radius values of the 6 wks-t group were between those of the control (13.2%; P>0.05) and 1 wk-t (−3.8%; P>0.05) groups (Fig. 5). Interestingly, the BM thickness tended to be lower (−12.7%; P=0.09, P=0.04 in Student's t-test) in the 1 wk-t group and was lower (−16.5%; P<0.05) in the 6 wks-t group than the control group, suggesting that the running wheel training was accompanied by a continuous thinning of the pericapillary BM in absolute size.

Some structural indicators were semi-quantitatively analyzed. The PC coverage at the abluminal EC surface differed non-significantly between the mice of the three study groups (control versus 1 wk-t: −1.4%; control versus 6 wks-t: −3.0%; 1 wk-t versus 6 wks-t: −4.3%). The relative enlargement of the intraluminal EC perimeters by protrusions was lower (−21% after 1 wk-t, −24% after 6 wks-t; P<0.05) in the PLNT muscle capillaries after the running wheel training than in the capillaries from the PLNT muscle of the control animals. The percentage of PLNT muscle capillary profiles with peg–socket junctions differed only non-significantly between the mice from the three study groups (EC sockets: control versus 1 wk-t: 20.7± 2.8%; control versus 6 wks-t: 19.4±6.5%; 1 wk-t versus 6 wks-t: 16.7±6.7%; PC sockets: control versus 1 wk-t: 3.0±4.5%; control versus 6 wks-t: 6.7±6.1%; 1 wk-t versus 6 wks-t: 2.5±4.2%).

Remarkably, the values for profile area sizes, compartment perimeters and profile area densities of the capillaries from the PLNT muscle of 6 wks-t mice were between those of the control and 1 wk-t groups (Table 1), suggesting that these changes in capillary structure established in the early phase reversed during the late phase of the training period. The coefficient of variation (CV) for all structural indicators differed 8.1–26.8%, being in the range of 20% for most indicators (data not shown).

Some structural peculiarities were discovered in capillaries depicted on the electron micrographs, which we describe here only qualitatively due to the low frequency of their occurrence (Fig. 6). Occasionally (N=4 from 360 capillary profiles), capillary profiles exhibited a second small lumen besides the major lumen (Fig. 6A). These branches might embody abluminal sprouts (characteristic of sprouting angiogenesis) but might alternatively represent tangentially sectioned or commencing branches of an established capillary. In some capillaries of the PLNT muscle from mice of the three study groups, one or more very long EC protrusions projected into the capillary lumen (Fig. 6B). The proportion of capillary profiles with intraluminal protrusions/filopodia longer than 5.2 μm (which would be able to divide a capillary lumen into two approximately equal-sized openings if they were connected to the opposite capillary wall) differed significantly between the controls and the two exercise groups (controls: 19.4±11.4%; 1 wk-t: 7.0±4.5%; 6 wks-t: 2.0± 6.1%). Once only, we noticed a clearly transversely sectioned muscle fiber, which was accompanied by an exactly orthogonally running capillary (Fig. 6C). Also singularly, a mysterious feature was seen on a micrograph, which could not be identified without doubt and probably represents a structural artifact generated by tissue shrinkage during the glutaraldehyde fixation (Fig. 6D).

In this investigation, we have characterized the running activity of C57BL/6 mice that were subjected to voluntary running wheel training for 1 week (1 wk-t) or 6 weeks (6 wks-t) in comparison to those of untrained littermates to subsequently assess morphometrically the capillarity and the ultrastructure of capillaries in the PLNT muscle of all mice. Essentially, we have made three major observations. (1) The mean daily running distance of the 6 wks-t mice significantly increased after the initial training week for an additional two weeks to establish a high-level equilibrium for the remaining three weeks of the running wheel training. (2) The higher C/F ratio in PLNT muscle observed in the 6 wks-t group compared with the controls was accompanied by a higher c(K,0) and a higher J_v. (3) The morphometric analysis of transmission electron micrographs revealed a tendency for lumen expansion of the capillaries in the 1 wk-t group but not in the 6 wks-t mice. The running wheel training of the mice was also accompanied by a continuous decrease in the CBMT and shortening of intraluminal protrusions/filopodia.

The running activity of the 6 wks-t mice was monitored throughout the training period. The statistical comparison revealed the daily running distance to be significantly increased only in the two weeks after the first training week. Thereafter, the mean daily running distance changed only non-significantly from week to week. Two other studies also report that C57BL/6 mice undergoing voluntary running wheel training showed increased activity for several weeks before persisting on a high level. In one study, the mice had already reached their maximum after 2 weeks (Waters et al., 2004), whereas the running distance increased over a time period of 4 weeks in the other study (Olenich et al., 2013). Although the reasons for the slightly varying kinetics of the running activity described in these studies are not known, it is possible that discrepancies in the impeller diameters (Waters et al., 2004: 11 cm; Olenich et al., 2013: 11.5 cm; our study: 18 cm) or differences in age of the mice when starting with training (Waters et al., 2004: 8 weeks; Olenich et al., 2013: not specified; our study: 12 weeks) contributed to these variations. However, all three studies are in agreement that mice cannot permanently improve their daily running performance but stabilize at a high level after several weeks of training. The two time points at which we collected muscle samples for structural analysis reflect this two-part kinetics: one group was derived from the phase of increasing running distance (after the first week of wheel training), whereas the second group originated from the equilibrium phase of running activity (after six weeks of wheel training).

The C/F ratio represents the most established structural indicator of capillarity in skeletal muscle, which is particularly used to provide experimental evidence for the occurrence of angiogenesis in this tissue (Hudlicka, 1998). The fact that the C/F ratio in PLNT muscle of the 6 wks-t group was significantly higher than in the control group indicates that physiological angiogenesis occurred in this period in response to long-term endurance exercise as previously demonstrated in humans (Andersen and Henriksson, 1977; Hoppeler et al., 1985) and rats (Olfert et al., 2001). Thus, endurance exercise is an effective trigger of angiogenesis (Egginton, 2009; Prior et al., 2004; Yan et al., 2011). As the C/F ratio was higher in the 1 wk-t group than the controls and in the 6 wks-t group than the 1 wk-t group, it can be assumed that the angiogenic process was continuously enduring throughout the training.

Because the C/F ratio is determined on transverse muscle sections, this indicator is representative only of the two-dimensional capillary arrangement but does not provide quantitative information about the isotropic (spatial) course of capillaries, e.g. caused by meandering capillaries with many anastomoses and/or branches. In contrast, the J_v, which represents an estimate of the total length of the capillaries within a defined tissue volume, is suitable to quantitatively describe the three-dimensional arrangement of the capillary system. For the evaluation of J_v, isotropic uniform random (IUR) sampling/sectioning according to classical stereological rules is formally the method of choice (Weibel, 1979). However, IUR on skeletal muscle is laborious to implement (Vock et al., 1996), because this stereological approach requires a large number of tissue sections that are not always available when analyzing muscle samples. Therefore, the dimensionless c(K,0) was introduced as an alternative to the analysis of IUR sections (Mathieu et al., 1983), which is calculated by relating the capillary density on transverse sections Q_A(0) to that on longitudinal sections Q_A(π/2). We would like to emphasize that discrepancies in the capillary tortuosity do not affect the C/F ratio to a significant extent. It should also be noted that other alternative methods for the estimation of capillary tortuosity have been developed (Gueugneau et al., 2016; Vincent et al., 2010), some of which are more laborious to carry out (Charifi et al., 2004; Janáček et al., 2011) than the protocol provided by Mathieu et al. (1983), which we have used in the present study. c(K,0) is characteristic of any muscle, species and preparation method (Mathieu-Costello et al., 1989) but is not significantly affected by parameters such as body size, aerobic capacity and hypoxia (Mathieu-Costello et al., 1989) as well as by endurance exercise in the oxidative soleus muscle of rats (Poole and Mathieu-Costello, 1989). However, c(K,0) is related to the sarcomere length in the muscle (Mathieu-Costello, 1987).

In our study, c(K,0) was significantly higher in the PLNT muscle of the 6 wks-t group than the control, indicating that the three-dimensional arrangement of the capillary network in this glycolytic muscle has changed during the training period by becoming more convoluted. We suggest that the increase in capillary tortuosity extends the diffusion capacity of oxygen/carbon dioxide and energy substrates, thereby contributing to the fiber shifting towards a more oxidative phenotype induced by endurance training (Freyssenet, 2007; Hood et al., 2006). Consistent with this hypothesis, an increase in capillary tortuosity was found to be related to the activity of oxidative enzymes in skeletal muscle fibers after 14 weeks of moderate ergometer training (Charifi et al., 2004). In addition, a computer simulation revealed that an increase in capillary tortuosity in skeletal muscle causes a higher tissue oxygenation, particularly when combined with anastomoses (Goldman and Popel, 2000). However, the mechanisms as to how endurance exercise triggers an increase in capillary tortuosity are not known. It has previously been speculated (Egginton et al., 2001) that the high rates of EC stretching during training result in elongation of the capillaries, so that they meander, as it has been observed by intravital microscopy (Ellis et al., 1990).

To the best of our knowledge, this is the first study in which the ultrastructural phenotype of capillaries in skeletal muscle of mice has been evaluated after endurance exercise. Because the PLNT muscle of the mice were prepared one day after the last training session, the ultrastructural changes of the here described capillaries were not acutely caused by the higher contractility but represent chronic adjustments in the capillary phenotype instead.

The profile area sizes associated with the capillary lumen in the PLNT muscle tended to be larger in the 1 wk-t mice than in the controls. Correspondingly, the luminal capillary radius tended to be higher in the 1 wk-t than the untrained mice. These findings indicate that the running wheel training resulted in the expansion of the capillary lumen after the first week of training, which could be due to exercise-induced higher cardiac output that increases blood flow through the capillaries in the periphery (Hellsten and Nyberg, 2016). After six weeks of training, the profile area size of the capillary lumen was again lower (but did not reach the baseline values). Obviously, the lumen-related structural adaptation of the capillary was reversed after the long-lasting training stimulus. It is therefore tempting to speculate that the blood flow is better distributed through the capillaries in the PLNT muscle after angiogenesis has occurred, which in turn reduces the wall stress in capillaries, resulting in a lower lumen diameter (Masuda et al., 2003).

The CBMT was significantly reduced as well as the sizes and numbers of the intraluminal protrusion were lower in the 1 wk-t group than in the control group and then again lower in the 6 wks-t group. These findings suggest that the running wheel training of the mice was accompanied by a continuous thinning of the CBMT and a reduction in the intraluminal protrusion surface of their PLNT muscle capillaries over time. A decrease of the CBMT in skeletal muscle after endurance exercise of humans was likewise observed in other studies (Baum and Bigler, 2016; Williamson et al., 1996). However, several potential triggers and causes for the increase of the CBMT, such as increased hydrostatic pressure, reduction in blood flow, more glycation events and chronic inflammation have been identified (Baum and Bigler, 2016). Whether the exercise-induced change(s) in extent of one or more of these triggers of CBMT thickening contribute(s) to the reversible response observed in this study, meaning the CBMT thinning, is an interesting issue that should be investigated in further studies.

Because it appeared likely to us that endurance exercise causes only temporal changes in the capillary ultrastructure, we have tested for statistical significances of our measurements by both formally correct ANOVA as well as pairwise Student's t-test. The calculations showed that a few structural indicators only tended to vary between the study groups when applying ANOVA, whereas they significantly differed in the Student's t-test: the capillary (profile) density N_A(c,f) between the 1 wk-t and 6 wks-t groups, the area size of the capillary lumen between the controls and both the 1 wk-t and the 6 wks-t groups, and the CBMT between the controls and 1 wk-t group. It is currently not possible to decide which of these differences in capillary structure are actually significant because they represent reversal adaptations of the microvasculature.

Taken together, the training-dependent changes in lumen and BM appearance in murine skeletal muscle capillaries described in this study are consistent with those observed in human skeletal muscle after endurance exercise. In contrast, increase of PC coverage was only observed in skeletal muscle capillaries of humans (Baum et al., 2015) but not in those of mice as shown here. Whether this distinction represents a species-specific difference in the structural adaptation of capillaries to exercise is an open question.

In skeletal muscle of rodents, several features of changes in the capillary phenotype characteristic of splitting and sprouting angiogenesis have been identified (Egginton, 2009; Egginton et al., 2001; Hudlicka, 1998). During splitting angiogenesis, a higher proportion of intraluminal irregularities, projections and septa combined with extensive cytoplasmic vacuolization of ECs were observed in skeletal muscle capillaries of prazosin-treated rats compared with those of control animals (Egginton et al., 2016; Zhou et al., 1998a). Sprouting angiogenesis in skeletal muscle of rats induced by surgical extirpation of a synergistic muscle was associated with an increase of abluminal EC processes, a higher PC coverage of capillaries, higher rates of EC mitosis and focal breakage of the pericapillary BM (Hudlicka, 1998; Zhou et al., 1998b). If these structural hallmarks for splitting and sprouting angiogenesis (prazosin model, overload model) are compared with the morphometric findings described in the present investigation, a dissenting picture emerges. Neither of the splitting nor sprouting angiogenesis-related findings mentioned above were observed in our study. Thus, it is yet not possible to make a statement about the mode by which physiological angiogenesis is realized in skeletal muscles of mice in response to endurance exercise (Yan et al., 2011).

We are aware that some methodological limitations may restrict the significance of our findings. (1) Although we are not able to make a statement about the running activities of the trained mice and their untrained control littermates, we consider this issue to be negligible due to the long recorded distance that the mice have trained on the running wheel. (2) We cannot exclude a technical bias during tissue treatment (e.g. shrinkage by glutaraldehyde fixation) but like to underline that all samples/sections were treated in the same way. However, it should be borne in mind that the values for the structural indicators presented here are not to be considered absolutely. (3) The capillary phenotype, even within a defined muscle, is highly variable, e.g. most structural indicators show CVs of about 20%. Thus, it is necessary to include a sufficient number of capillary profiles in the morphometric analysis. We used 360 micrographs of capillaries in our study (120 per group), which appears to be a sufficient number, whereas a number of 6–17 capillaries per study group is certainly too low (Uchida et al., 2015) and may lead to wrong conclusions and interpretations of the outcome. (4) Arithmetic values as provided in this study represent only structural estimates. For a more functional interpretation of the morphometric findings, such as their potential relationship to oxygen and substrate supply, other indicators are more appropriate, e.g. the harmonic mean barrier thickness (Weibel, 1979), which takes into account the fact that thinner segments contribute more to diffusion than thicker ones in a proportional fashion.

In conclusion, our morphometric study performed at the light microscopy and electron microscopy levels revealed both the capillarity and the capillary ultrastructure in PLNT muscle to change over time during long-term endurance exercise training. In the early phase of the training period, angiogenesis and a tendency of capillary lumen expansion was accompanied by a significant reduction in CBMT and a shortening of mean intraluminal protrusion length and number. After long-term training, when the mice reached a steady state in running activity, additional angiogenesis and an increase in capillary tortuosity was established, which was accompanied by a partial reversal of the lumen expansion as well as further reductions in CBMT and shortening of the intraluminal protrusion length. The knowledge of these non-designed structural adjustments in the capillary phenotype may support the understanding of the changes in functionality of the microvasculature in response to endurance exercise, especially if these training-induced microvascular remodeling manifestations are regarded as control parameters in the negative feedback control circuit.

We would like to thank Christoph Lehmann, Bern, very much for his excellent technical support in assembling and maintaining the running wheel equipment. Thanks to Anna Stocker for proofreading the manuscript. The assistance of Barbara Krieger to improve the graphical presentation is likewise highly appreciated.