Maximum aerobic performance in lines of Mus selected for high wheel-running activity: effects of selection, oxygen availability and the mini-muscle phenotype

The vertebrate O₂ transport and delivery system can be considered to be composed of four steps in series: ventilatory convection,alveolar-capillary diffusion, blood convection, and tissue capillary-to-cell diffusion of O₂ (Taylor and Weibel, 1981; Wagner,1996). How whole-animal maximum aerobic capacity(V̇_O2max) should respond to changes in ambient partial pressure of O₂(P_O2) depends on which of these steps is limiting and on mitochondrial oxidative capacity in peripheral tissues. For instance, hypoxia should cause a reduction in V̇_O2max unless both O₂ delivery and mitochondrial oxidative capacity were in excess. In this context, Lindstedt et al.(1988) have shown that when O₂ delivery is reduced by anemia or hypoxia, V̇_O2max declines in direct proportion to O₂ delivery, whereas O₂extraction from the blood remains relatively constant (∼90%). Effects of hyperoxia on V̇_O2max are not as straightforward, because limitations at different levels in the O₂cascade may lead to different experimental outcomes(Richardson et al., 1999; Lindstedt and Conley, 2001; Noakes et al., 2001). Increased V̇_O2maxin hyperoxia may suggest that O₂ uptake and delivery systems are more relevant for V̇_O2max in normoxia than mitochondrial oxidative capacity (i.e. central limitation),whereas potential constraints at the mitochondrial level in the muscles might be the case if V̇_O2max remains unchanged at higher P_O2 (i.e. peripheral limitation).[However, limits in cardiac output or blood O₂ saturation in the latter example could not be ruled out, as convective steps in the O₂ cascade (e.g. O₂ movement into the lungs, blood O₂ transport, muscle blood flow) might impose a ceiling in V̇_O2max regardless of the higher diffusion consequent of the increase O₂gradient.]

Alternatively, it could be the case that all steps in the O₂cascade are virtually identical in capacity, such that none, by itself, would set the limit for maximum organismal performance. Indeed, symmorphosis(Taylor and Weibel, 1981) has been proposed as a response at many levels of biological organization to the`powerful optimizing process'(Alexander, 1989, p. 1200)attributed to natural selection. Whether animals are symmorphotic presently remains a matter of debate among physiologists(Hammond and Diamond, 1997; Bacigalupe and Bozinovic,2002), and several authors have argued against its biological relevance in the light of evolutionary theory and various empirical results(Garland and Huey, 1987; Garland, 1998; Gordon, 1998; brief review in Suarez and Darveau, 2005). Almost all existing studies aimed at testing symmorphosis have been interspecific comparisons, so experimental evolution can offer a novel approach to the study of such hypotheses about the correlated evolution of complex phenotypes (Garland, 2001, 2003; Swallow and Garland, 2005; Garland and Carter, 1994).

Artificial selection for high voluntary wheel running in four replicate lines of house mice (Swallow et al.,1998a) provides an opportunity to examine interactions between aerobic limits and locomotor activity. Existing evidence indicates that normoxic V̇_O2maxhas increased, especially in males, coincident with the evolution of higher activity levels in the selected (S) lines as compared with their random bred control (C) lines (Swallow et al.,1998b; Rezende et al., 2005, 2006). Therefore, selection might have affected hypoxia tolerance at the muscle level because these lines run voluntarily at speeds approaching their maximum aerobic speeds for at least some minutes per night (Girard et al., 2001; Rezende et al., 2005, 2006).

An unexpected part of the response to selection has been the evolution of the `mighty mini-muscle' phenotype, which represents homozygotes for a Mendelian recessive allele that halves gastrocnemius muscle mass(Garland et al., 2002; Belter et al., 2004) while doubling per gram aerobic capacity(Houle-Leroy et al., 2003). This phenotype occurs in only two of the four selected lines and has now gone to fixation in one of them. Apparently, the other two selected lines lost the allele, which was initially rare, by random genetic drift during the early generations of the experiment (Garland et al., 2002). In any case, the increase in frequency of this gene of major phenotypic effect represents an important component of the overall response to selective breeding. Moreover, because it has significant effects on a variety of other traits in addition to gastrocnemius muscle mass(Garland et al., 2002; Houle-Leroy et al., 2003; Swallow et al., 2005; Syme et al., 2005; Kelly et al., in press), it should be accounted for in statistical analyses. For example, the soleus muscle of mini-muscle individuals is actually larger than in normal individuals (Syme et al.,2005). And, considering that mini-muscle individuals often run both more and faster on wheels as compared with normal-muscled individuals within the same line (Syme et al.,2005; Kelly et al., in press), one might expect that the phenotype could also affect V̇_O2max.

The purpose of this study was to determine the effects of different atmospheric P_O2 on V̇_O2max during forced exercise in the S and C lines of mice and to test for pleiotropic effects of the mini-muscle phenotype. Results can be important for several reasons. First, they may provide insight about the physiology underlying V̇_O2max and the relative importance of O₂ uptake and delivery systems at the whole-individual level. Second, given that S lines have elevated V̇_O2max as compared with C lines (Swallow et al.,1998b; Rezende et al., 2005, 2006), hypoxia and hyperoxia may affect S and C lines differently, which would suggest differences at lower physiological levels in the O₂ cascade, as has been found in lines of rats selected for high vs low treadmill performance (e.g. Henderson et al., 2002; Howlett et al., 2003). Such results would be relevant for symmorphosis. Specifically, if S lines have evolved in a symmorphotic fashion, one would expect effects of P_O2 on V̇_O2max to be independent of selection history, and differences between S and C would remain constant across different P_O2. Conversely, if S and C lines respond differently to changes in P_O2, this would suggest that traits subordinate to V̇_O2max have evolved in a `non-symmorphotic' fashion in response to selection. Third,effects of hyperoxia on V̇_O2max and running performance could reveal whether aerobic constraints may be limiting to the evolution of even higher levels of wheel running in the S lines. An increased performance by S mice in a hyperoxic atmosphere would suggest that they might run even more (faster) on wheels if O₂ availability at the tissue level were higher. Finally, results may help elucidate why the mini-muscle phenotype has been favored by the selection regimen(Garland et al., 2002; Houle-Leroy et al., 2003; Swallow et al., 2005; Syme et al., 2005; Kelly et al., in press).

We used 59 females from generation 36 post-selection of a selection experiment for high voluntary wheel running, as described in detail elsewhere(Swallow et al., 1998a; Garland, 2003). Briefly,voluntary running performance was measured on wheels every generation in eight independent lines of laboratory mice (∼10 families per line each generation). After founding of the initial colony, four of the lines were randomly assigned to the selection treatment (lab designations 3, 6, 7 and 8)and the other four were treated as controls (lines 1, 2, 4 and 5). In the four selected (S) lines, the best male and female runners within each family were used as breeders for the next generation (and second-highest runners as additional extra breeders, if necessary). The selection criterion was running distance during the last 2 days of a 6-day trial. In the four control (C)lines, breeders were randomly chosen at weaning (21 days of age). Selection was interrupted for four generations (32-35) to transfer the colony from the University of Wisconsin-Madison to California. For generation 36, a sample of which was studied here, the average difference between S and C mice in wheel running on days 5+6 of the 6-day trial was 2.37-fold (averaged over all eight lines and both sexes; T. Garland, Jr., unpublished results; P_selection<0.001).

After breeders for the next generation were selected, we obtained our individual females from different families from within each line. We used six females per line, except for line 6, which is polymorphic for the mini-muscle allele mentioned above (sibs were sometimes used in this line; see Statistics), where we used 17 females. Note that line 3 is now apparently fixed for the mini-muscle allele (i.e. all individuals have the phenotype; Syme et al., 2005; T. Garland,Jr., unpublished results). Prior to choosing these individuals for study, we excluded the lowest runner within each family to compensate for the bias caused by having chosen the highest runners to become breeders during the selection protocol (see previous paragraph). Mice were housed four per cage from weaning to selection, with food and water ad libitum, then measured for 6 days on wheels, beginning at around 62 days of age (range 54-70 days), and subsequently were maintained in individual cages.

At approximately 95 days of age (79-110 days), individuals were randomly assigned to five measurement batches of 12 mice each (although within each line, older individuals were assigned to initial batches to minimize differences in age). Each batch was used for treadmill measurements at three different P_O2 (see below) for six consecutive days -measurements were performed twice on consecutive days at each P_O2. To control for training or `acclimating' effects,batches were measured in each atmosphere in different sequences, and individuals were measured in random order within each batch. All trials were performed between 09.00 and 12.00 h. On day 7, mice were sacrificed and dissected, and individuals with the mini-muscle phenotype were identified by inspection of graphs of muscle mass versus body mass (see Garland et al., 2002; Houle-Leroy et al., 2003; Belter et al., 2004; Syme et al., 2005).

Maximum aerobic capacity during forced exercise(V̇_O2max) was estimated with open-flow respirometry by running mice in an enclosed motorized treadmill, as described previously (Hayes and Chappell, 1990; Chappell et al., 2003; Rezende et al.,2005). The treadmill had an inclination of 25°, which maximizes the V̇_O2max values obtained in mice (Kemi et al.,2002) and resulted in significantly higher V̇_O2max in males from generation 33 when compared with males tested without slopes (E.L.R.,T.G., F.R.G. and M.A.C., unpublished data). Treadmill tests were performed at room temperature (22-25°C). We used positive-pressure, flow-through respirometry to determine rates of oxygen consumption(V̇_O2) and carbon dioxide production(V̇_CO2). Airflow was regulated with upstream mass flow controllers (Applied Materials, Tylan,Billerica, MA, USA), with a flow rate of 2100 ml min^-1, that maintained excurrent O₂ concentrations within 0.5% of incurrent gas. Approximately 100 ml min^-1 of excurrent air was subsampled and analyzed for CO₂ (which was then scrubbed) and O₂. Oxygen concentration was measured with an Ametek/Applied Electrochemistry S-3A analyzer (Pittsburgh, PA, USA), and CO₂ was measured with LiCor 6251. Data from gas analyzers and other instruments were recorded on Macintosh computers equipped with National Instruments A-D converters and `Labhelper'software (Warthog Systems, www.warthog.ucr.edu).

To maintain different atmospheric O₂ concentrations, flow controllers were supplied with room air or connected to cylinders with different mixtures of oxygen and nitrogen creating hypoxia (14% O₂)and hyperoxia (30% O₂). Reference readings of incurrent gas were obtained at the start and end of measurements. References from cylinders were always checked against room air before measurements and remained precise(±0.1%) throughout the experiment. Mice were placed inside the treadmill's working section (6 cm wide, 7 cm high, 13.5 cm long), which was flushed for ∼1 min to attain stable O₂ concentration during hypoxia and hyperoxia (for consistency, this was also applied in normoxic trials). Mice were allowed a 1-2 min adjustment period and were then run at increasing speeds starting at 0.15-0.2 m s^-1 and raised in step increments of ∼0.15 m s^-1 every 45 s. A trial was terminated when mice could no longer keep pace with the treadmill and V̇_O2 did not increase with increasing tread speed. All animals failed to maintain position at the highest tread speeds implemented. Trial quality was also assessed using a subjective scale (five categories, from `poor' to `excellent'; following Swallow et al., 1998b; Rezende et al., 2005). In this sample of mice, no individual was scored as `poor' and hence none were excluded from analyses. Preliminary analyses using trial quality as an additional covariate indicated that it was never statistically significant, so it was not included in final models.

where F_ICO2 and F_ECO2 are the fractional CO₂ concentrations of incurrent and excurrent gas, respectively(F_ICO2 was ∼0.0004 in normoxia and 0 in hypoxia and hyperoxia). We determined V̇_O2max and V̇_CO2max as the highest 60-s continuous average in the trial, and calculated the respiratory exchange ratio(RER=V̇_CO2/V̇_O2)at V̇_O2max and V̇_CO2max. Maximum running speed attained in each trial was measured with a calibrated tachometer attached to the treadmill's motor drive. All calculations and corrections on V̇_O2 files were performed using LabAnalyst software (Warthog Systems).

Analyses were performed with SPSS for Windows and SAS PROC MIXED (version 8, SAS Institute, Cary, NC, USA, 1996). Repeatability between traits, measured on two consecutive days, was estimated with Pearson product-moment correlations (Hayes and Jenkins,1997) on residuals from nested analysis of covariance (ANCOVA),correcting for selection history (line type and line), mass, batch and age(see below). To check for day-to-day `training' effects, we used paired t-tests comparing trials 1 and 2.

As shown below, individual performance in the first and second trials within a given P_O2 was highly repeatable, consistent with maximal effort. Because we were interested in maximum performance, we selected the higher of the two V̇_O2max recorded at each P_O2 for most analyses, consistent with previous studies (e.g. Swallow et al.,1998b; Rezende et al.,2005). Differences between line types (S versus C) were assessed with type III sums of squares, using linear mixed models. Line type was the grouping variable, replicate lines (N=8) were nested within line type as a random factor, mass and age were covariates, and batch was a cofactor. Line effects were determined by comparing maximum-likelihood (ML)estimates with and without lines in the model (the difference between ML values follows a χ² distribution with 1 d.f.). Least-squares means adjusted to a common age and body mass were calculated to estimate the difference between S and C lines. Regular ANCOVAs were also performed for S and C separately, including line as a random factor (four lines) and using the same covariates of the nested ANCOVA. Discriminating effects of the mini-muscle allele from other effects is not straightforward (see Discussion);hence, we employed two different approaches. First, nested ANCOVAs were performed with an additional dummy variable equal to 0 for normal phenotypes and 1 for mini-muscles (N=48 and 11, respectively). Second, we compared individuals with and without the mini-muscle phenotype within line 6,employing regular ANCOVA with mass and age as covariates.

We assessed how V̇_O2 changed as a function of O₂ concentration and whether these changes differed due to selection history, employing general linear models (GLM) for repeated measures (SPSS for Windows). Individuals were experimental units, partial pressure (hypoxia, normoxia and hyperoxia) was the within-subjects factor, and selection history and lines were included as between-subjects factors (the latter, in this case, was treated as a fixed factor). Significance of selection history (S versus C) was estimated by dividing mean square error (MSE) of the model including only line type as the fixed factor by MSE from the model with line × line type (i.e. same d.f. obtained in the nested ANOVA). To determine how variables differed between trials, contrasts(differences between successive values for each individual) were compared employing multivariate ANOVAs (test of within-subject contrasts). We also tested for individual effects across different P_O2 with an F ratio between the MSEs between- and within-individuals (=MSE_between/MSE_within), obtained from a repeated-measures ANOVA with individuals as the grouping factors. When there are only two repeated measures, this test is analogous to Pearson's correlations and can also be used to test for `consistency' between measurements, i.e. `repeatability' (Hayes and Jenkins, 1997; Zar,1999; Rezende et al.,2005).

All P values shown correspond to two-tailed tests, although in many cases one-tailed tests would be appropriate. For example, in all measures of repeatability of individual differences, the alternative hypothesis is for a positive correlation, not just a correlation different from zero (unless fatigue effects were large and carried over for 24 h). Similarly, for V̇_O2max, the expectation from both first principles and previous studies(Swallow et al., 1998b; Rezende et al., 2005) is that mice from S lines will have higher values than those from C lines. On the other hand, multiple tests are performed that compare the same set of animals for various traits, which would suggest some sort of correction for multiple comparisons (e.g. see Rice,1989; Curran-Everett,2000). Thus, we report test statistics, d.f. and P values for all statistical tests; independent conclusions may be drawn from these as desired.

Maximum rates of O₂ consumption(V̇_O2max) and CO₂ production(V̇_CO2max) were highly repeatable in hypoxia, normoxia and hyperoxia, regardless of selection history (Table 1). Respiratory exchange ratios obtained at V̇_O2max(RER_V̇O2max)were repeatable only in hypoxia and normoxia (P<0.02 in all cases), whereas RER_V̇CO2maxwere not repeatable in most analyses. Maximum running speeds (treadmill with 25° slope) in S and C lines were always significantly repeatable between days (Table 1). Accordingly,effects of individual variation across P_O2 remained statistically significant for V̇_O2max, V̇_CO2max and maximum running speeds but not for either of the RER indexes (see below).

Training effects were negligible for V̇_O2max, which was 2.2% lower in the first trial in hypoxia (two-tailed P=0.023), and the same was true for V̇_CO2max(P=0.003). No differences between trials 1 and 2 were observed in maximum running speeds at any P_O2 (P>0.05 in all cases).

Mean (unadjusted) body mass was 26.6±0.3 g (mean ± s.e.m.)for S lines and 30.4±0.5 g for C lines. Unadjusted individual V̇_O2max was 4.888±0.083 ml O₂ min^-1 and 4.614± 0.079 ml O₂ min^-1 for S and C lines in hypoxia (i.e. 0.184±0.003 ml O₂ min^-1 g^-1 and 0.152±0.003 ml O₂ min^-1 g^-1 on a per gram basis), 5.662±0.124 ml O₂ min^-1 and 5.225± 0.130 ml O₂ min^-1 in normoxia(0.212±0.005 ml O₂ min^-1 g^-1 and 0.171±0.004 ml O₂ min^-1 g^-1) and 6.104±0.138 ml O₂ min^-1 and 5.481±0.151 ml O₂ min^-1 in hyperoxia (0.229± 0.005 ml O₂ min^-1 g^-1 and 0.182±0.005 ml O₂ min^-1 g^-1).

According to the nested ANCOVA, mice from S lines were significantly smaller in body mass than C mice (adjusted means of 27.6±1.5 g and 31.5±1.5 g, respectively, P_selection=0.0038; Fig. 1), but there were no differences between normal and mini-muscle phenotypes (full model nested ANCOVA, P_mini=0.720). In addition, according to regular ANCOVA within line 6 with age as covariate, body mass did not differ between mini-muscle and normal mice (F_1,14=0.088, P=0.771).

After accounting for variation in body mass, S lines achieved higher V̇_O2max and V̇_CO2max than C mice (Fig. 2), with the greatest difference in hyperoxia (Table 2). In the full nested model, individuals with mini-muscles had statistically higher V̇_O2max and V̇_CO2max only in hypoxic conditions (Table 2; Fig. 1). Considering only individuals within selection line 6, those with the mini-muscle phenotype achieved statistically higher V̇_O2max and V̇_CO2max in hypoxia (Table 3) but not in normoxia or hyperoxia (P>0.21 in all cases; Table 3).

Maximum running speeds were significantly higher in S mice at all O₂ concentrations (Table 2; Fig. 2). Regardless of atmospheric P_O2, V̇_O2max and V̇_CO2max increased significantly with body mass, whereas maximum running speeds did not(Table 2). Speeds were not affected by mini-muscle status in the full data set(Table 2) or within line 6(Table 3).

Repeated-measures analyses show that increased P_O2 led to increased V̇_O2max and V̇_CO2max in all lines, regardless of selection history(Fig. 2; Table 4). Similarly,within-subject contrasts - i.e. testing whether the mean at a given P_O2 is significantly different from the mean obtained at the previous level - show that V̇_O2max mean values were always significantly higher at higher P_O2,regardless of selection history (F_1,51=74.24, P<0.0001 for hypoxia versus normoxia, and F_1,51=57.35, P<0.0001 for hyperoxia versus normoxia; Fig. 2A). The same was true for V̇_CO2max(P<0.008 in both cases; Fig. 2B). Selection history had a marginal effect on V̇_O2max(one-tailed P=0.057; Table 4), supporting the conclusion that S mice attain higher V̇_O2max at any given P_O2 (Fig. 2A). This was not the case with V̇_CO2max(one-tailed P=0.237).

Changes in P_O2 had a less dramatic effect on V̇_CO2max than on V̇_O2max(Fig. 2). Accordingly, RER at V̇_O2max and V̇_CO2max changed significantly between ambient P_O2. In addition, both RER indexes were significantly lower for S mice when mass effects were not controlled (Tables 2, 4). Contrast analyses from repeated-measures ANCOVA found significant differences in RER between hypoxia versus normoxia, and normoxia versus hyperoxia(P<0.0001 in both cases). In addition, maximum running speeds on the treadmill also changed significantly with P_O2, being 13% higher in normoxia than in hypoxia (F_1,57=54.78, P<0.001) and 4.2% higher in hyperoxia than in normoxia(F_1,57=48.73, P<0.001; Fig. 2D). Line type (S versus C) was statistically significant as between-subject effects(F_1,6=76.43, P<0.001), indicating that S mice achieved higher speeds. Comparisons among contrasts were not performed separately for S and C because the line type × trial interaction was never statistically significant (i.e. responses across trials did not differ between S and C).

Values of V̇_O2max at different P_O2 were significantly correlated at the level of individual variation after accounting for selection history, mass, age and batch (Fig. 3). Correlations performed separately for S and C, employing residuals from regular ANCOVA with lines as grouping factors, provide identical results: individual consistency in V̇_O2max across P_O2 was statistically significant regardless of selection history (r>0.352, P<0.046 in all cases; Fig. 3).

Previous studies of these same lines at generations 10, 32 and 34 also indicated that S mice have higher average aerobic capacities than C mice in normoxia (respectively, +6% for males in Swallow et al., 1998b; +18%for males in Rezende et al.,2006; +7% for females in Rezende et al., 2005). Although our a priori expectation was that mice from the S lines would evolve higher V̇_O2max, a small response, especially at earlier generations, is consistent with a study of the base population, which found limited evidence for narrow-sense heritability in this trait (Dohm et al.,2001). In the present study of females from generation 36,controlling statistically for body mass but not for the mini-muscle phenotype,S lines attained V̇_O2max averaging 13.2% higher than C in normoxia (one-tailed P=0.0294; Table 2; Fig. 2A), although the values are within the range of those previously reported(Swallow et al., 1998b;Rezende et al., 2005,2006). Accordingly, we also detected a marginal effect of line type on V̇_O2max in the repeated-measures ANOVA (one-tailed P=0.057; Table 4), supporting the conclusion that S females attained higher V̇_O2max at all P_O2 as compared with C mice(Fig. 3).

In hypoxia, S lines attained V̇_O2max averaging 10.7% (P=0.0291) higher than C, whereas in hyperoxia the difference was 20.8% (P=0.0074; Table 2; Fig. 2A). The repeated-measures analysis was unable to detect a significant line type× P_O2 interaction (two-tailed P=0.103; Table 4), although Table 2 and Fig. 2 suggest that S and C lines actually do respond differently when exposed to variation in P_O2, particularly between normoxia and hyperoxia. Moreover, quadratic effects in the repeated-measure analysis bordered significance in C lines (F_1,20=4.08, two-tailed P=0.057) but not in S lines (F_1,31=1.16, P=0.289), in spite of the higher sample size in the latter group. That is, within the range of P_O2 studied, V̇_O2max seems to increase essentially linearly in S mice but plateaus at normoxic levels in C mice (Fig. 2A).

With respect to maximal CO₂ production, Astorino and Robergs(2003, p. 13) stated that`an enhanced V̇_CO2may represent greater oxidative ATP production, yet few studies have demonstrated significant increases in V̇_CO2in hyperoxia'. Contrasts analyses show that S mice attained significantly higher V̇_CO2max in hyperoxia versus normoxia (F_1,31=7.45, P=0.010), but C mice did not (F_1,20=1.58, P=0.223; Fig. 2B),suggesting that oxidative metabolism increased significantly in S lines during trials in hyperoxia. Although V̇_CO2max could increase because of higher muscle fiber recruitment in hyperoxia, or increased oxidative metabolism, the significant reductions observed in RER in hyperoxia(Fig. 2C) support the latter alternative (see below). Consistent with this possibility, Houle-Leroy et al.(2000) studied both sexes at generation 14 and found that S mice housed without wheel access exhibited a trend for higher mixed hind limb muscle aerobic capacities, as indicated by higher levels of mitochondrial (cytochrome c oxidase, carnitine palmitoyltransferase, citrate synthase, pyruvate dehydrogenase) and glycolytic(hexokinase, phosphofructokinase) enzymes, with lower anaerobic capacities, as indicated by lactate dehydrogenase (especially in males). [In males from generation 10, when the S versus C differential in wheel running was substantially less, Zhan et al.(1999) found no evidence of elevated succinate dehydrogenase activity in the medial gastrocnemius muscles of S mice.] Interestingly, V̇_CO2max is related to some types of locomotor endurance at the level of individual variation in lizards (Garland,1984).

The mini-muscle allele is still segregating within line 6 but is now apparently fixed in line 3 (see also Syme et al., 2005). We therefore increased sample size within line 6(see Materials and methods) to allow within-line comparisons. Whether any effects within line 6 would be the same as in line 3 are unknown and difficult to study because these lines probably differ in allele frequencies at many other loci, leading to possible effects of different genetic backgrounds and/or epistatic interactions. Nevertheless, analyses of the entire data set(all eight lines) and just within the polymorphic S line 6 were consistent:individuals with the mini-muscle phenotype had significantly higher V̇_O2max and V̇_CO2max under hypoxic conditions but not in normoxia or hyperoxia (Tables 2, 3; Fig. 1). Although previous studies have described a twofold increase in mass-specific oxidative capacity in hind limb muscles of individuals with the mini-muscle phenotype(Houle-Leroy et al., 2003),and higher wheel-running distances and speeds in some samples(Syme et al., 2005; Kelly et al., in press), this is the first report of significant effects of the mini-muscle phenotype on aerobic capacity at the whole-organism level.

Although the most straightforward explanation for elevated V̇_O2max would be increased mass-specific aerobic capacity (in vitro catalytic rates of oxidative enzymes; Houle-Leroy et al.,2003), several lines of evidence suggest that additional factors may be involved. First, individuals with the mini-muscle phenotype have gastrocnemius almost 50% lighter than normal, compensating for the twofold increase in mass-specific oxidative capacity(Garland et al., 2002; Houle-Leroy et al., 2003; Syme et al., 2005). Second,one might expect that any favorable effect of the mini-muscle allele would be more pronounced under those conditions in which it was selected (normoxia). By contrast, differences between mini-muscle and normal individuals were somewhat inconsistent across different P_O2 (i.e. genotype-by-environment interaction; Tables 2, 3). Although this does not exclude the possibility that higher cellular oxidative capacities could increase O₂ gradient from capillaries to muscles, increasing overall O₂ extraction from the air and ultimately V̇_O2max under hypoxic conditions, that scenario assumes that effects of hypoxia on additional steps of the O₂ cascade are negligible, which might not be the case (see below). Third, pleiotropic effects of the mini-muscle allele on other steps in the O₂ cascade are possible (see also Belter et al., 2004; Swallow et al., 2005). Individuals with the mini-muscle must provide the same amount of O₂to their muscles as normal mice but are constrained by smaller overall muscle cross-sections, and potentially fewer capillaries, to transport blood into these muscles. Mice with the mini-muscle phenotype have significantly larger ventricle mass than normal (Garland et al., 2002; Swallow et al.,2005; unpublished result in Gomes et al., 2004), which might be indicative of compensations at other levels, such as higher cardiac outputs. Another possibility is that the capillary/fiber ratio has increased in the mini-muscle, as has been described in muscles with high oxidative capacity (Hepple and Vogell,2004; and references therein), which could facilitate the O₂ flux to muscles and decrease the necessity of increased blood pressures.

Factors limiting maximum O₂ uptake may be broadly broken into four categories: (1) pulmonary diffusing capacity, (2) maximum cardiac output,(3) oxygen transport in the blood and (4) skeletal muscle characteristics(Wagner, 1996; Richardson et al., 1999; Bassett and Howley, 2000; see Hochachka, 2003, for details of O₂ transport within cells). According to Fick's law of diffusion and the Fick principle, the relative importance of each of these factors in determining V̇_O2max may vary at different atmospheric P_O2. Hence, the interaction between environmental P_O2, selection history and the presence or absence of the mini-muscle phenotype and their effects on V̇_O2max must be considered in a holistic manner.

Our results show that neither S nor C lines can compensate for the lower P_O2 gradient during hypoxic trials(Fig. 2). Assuming a mixed venous P_O2 of 21 and 34 Torr (1 Torr=133.3 Pa) during V̇_O2max in hypoxia and normoxia (based on measurements performed in rats by Henderson et al., 2002; their inspired P_O2 in hypoxia was ∼30 Torr lower than ours,however) and everything else being equal, one would predict a decrease of 32%in V̇_O2max in hypoxia compared with normoxia, contrasting with 12.9±1.6% and 10.9±1.9% for S and C, respectively(Table 2; Fig. 2).Although lower diffusion rates are probably the main causal factor, several physiological processes, such as reduced cardiac output, might be contributing to lower V̇_O2max in hypoxia(e.g. Gonzalez et al., 1998; Calbet et al., 2003).

The mean increase of 7.2% in V̇_O2max in hyperoxia versus normoxia is considerably lower than the 55.1%increase predicted if venous P_O2 was assumed to be identical to values obtained in normoxia (above), supporting the hypothesis that muscle aerobic capacity might constrain higher V̇_O2max in hyperoxia. Accordingly, previous studies have reported relatively minor increases in V̇_O2max with increased O₂ delivery (Spriet et al., 1986; Lindstedt et al., 1988; Richardson et al.,1999; Lindstedt and Conley,2001), which suggests that O₂ delivery was enhanced beyond the capacity of mitochondria to metabolize the O₂ available. Although it is possible that mitochondrial capacity might be reaching its limit in 30% O₂ (e.g. Richardson et al., 1999 in humans at 100% P_O2), two lines of evidence suggest that V̇_O2max in S and C lines under normal conditions (i.e. normoxia) is constrained by pulmonary and cardiovascular systems. First, RER decreased significantly as P_O2 increased (Fig. 2C), which might be due to a higher O₂ transport,facilitated by higher O₂ concentration at the alveolar level. Second, both S and C mice had significantly higher V̇_O2max in hyperoxia compared with normoxia (see Results), emphasizing that peripheral muscles can increase aerobic metabolism when more intracellular O₂is available.

The differential response of V̇_O2max between S and C mice during hyperoxia suggests that selection for high wheel running affected differentially some of the components involved in the O₂cascade. Polynomial analyses suggest that C animals could not increase V̇_O2max in hyperoxia to the extent that S mice did(Fig. 2A), although the P_O2 × line type interaction was not statistically significant according to a two-tailed hypothesis(Table 4). Nevertheless, ANOVAs and ANCOVAs have relatively low power to detect interactions (e.g. Wahlsten, 1990; Houle-Leroy et al., 2000), and it is reasonable to suggest that V̇_O2max responses to hyperoxia differ between S and C lines(Table 4; P_O2 × line type F_2,12=2.77,one-tailed P=0.051).

If so, then two non-exclusive hypotheses are possible, assuming that O₂ transport within the cell and mitochondrial oxidative capacity do not limit cellular metabolism (see above) or differ between line types(e.g. Houle-Leroy et al.,2000). First, O₂ extraction in lungs and/or muscles might have increased in S lines. Henderson et al.(2002) showed that rats selected for high treadmill endurance during forced exercise evolved higher V̇_O2max despite minimal changes in convective O₂ transport (cardiac output was significantly different between lines, however; Hussain et al., 2001). A higher O₂ extraction was attributed to greater muscle fiber capillarization (smaller muscle fibers, same number of capillaries) in the high endurance performers (see also Howlett et al., 2003). If the same pattern is true in our lines, this could explain why S lines have consistently higher V̇_O2max than C mice regardless of P_O2.

If differences between S and C were simply caused by unequal O₂extraction, then one would expect the relationship between P_O2 and V̇_O2max to be linear for both S and C, but with different slopes. The presence of a significant quadratic component in C lines(Fig. 2A) suggests that other factors might be involved. Therefore, a second alternative is that O₂ convection (either due to ventilatory convection or cardiac output) limits higher V̇_O2max in C lines- but not in S lines - in normoxia and hyperoxia. One possibility is that hyperoxia might be preventing hemoglobin desaturation at high exercise levels in S lines, as reported for humans(Nummela et al., 2002). Breathing frequency was significantly higher in S lines during V̇_O2max in a He-O₂ atmosphere (Rezende et al., 2005), although it is not clear if that would be the case during forced exercise.

Alternatively, S mice might have increased cardiac output. There is general agreement that higher rates of blood flow result in an improvement in V̇_O2max (e.g. Saltin and Strange, 1992), and we have recently found that S mice have significantly larger ventricles after accounting for mini-muscle effects (Gomes et al., 2004; differential training effects between S and C may be a factor because mice were measured after 6 days of wheel access). Although severe hyperoxia (100% O₂) causes reduction of microvascular flow due to vasoconstriction in resting hamsters(Tsai et al., 2003), we do not know if any such effects might occur at 30% O₂ during strenuous exercise in mice, or if increased cardiac outputs and presumably higher blood pressures could prevent such maldistribution of perfusion in the microcirculation. Nevertheless, significantly lower RER in S lines(Table 4) suggests that selection for high activity has led to more efficient O₂ transport to muscles.

Inspection of the values in Table 3 shows that mice with mini-muscles always tended to have higher V̇_O2max than those with the normal phenotype (see also Fig. 1). Although larger hearts and higher mass-specific muscle aerobic capacity suggest that increases in cardiac output, blood pressure and P_O2 gradient at the capillary level might potentially explain (at least partially) the differences in aerobic capacity between phenotypes, further studies are required to address this question. Indeed,many of the factors that could explain S versus C differences could also explain differences between mini versus normal phenotypes (e.g. greater muscle fiber capillarization; as in Howlett et al., 2003).

Finally, some studies have proposed that central nervous system limitations might constrain V̇_O2max in hyperoxia (see Astorino and Robergs,2003). Neurobiological differences between S and C lines have been described, especially regarding dopaminergic function (review in Rhodes et al., 2005), and dopaminergic function is known to be involved in respiratory control in mice(e.g. Huey et al., 2003). The higher variability in V̇_CO2max estimates in normoxia in both S and C lines (Fig. 2B) might be related to changes in ventilatory regulation in this treatment, associated with the presence of atmospheric CO₂ in the incurrent gas - i.e. CO₂ was not removed in normoxia, and baseline measurements were used to obtain corrected values of V̇_CO2 (see Materials and methods). This possibility seems very unlikely, however, given the very small fraction of CO₂ inspired compared to lung P_CO2. It is also unclear whether ventilatory differences between S and C lines (above), possibly associated with changes in dopamine function per se, could explain the `plateau' in V̇_O2max observed in C lines in hyperoxia (Fig. 2A).

Our results support previous indications that selection for high voluntary wheel running has caused increased aerobic capacity(Swallow et al., 1998b; Rezende et al., 2005, in press). Although we have previously reported correlated responses at the level of peripheral tissues (e.g. increased frequency of the mini-muscle allele leading to more oxidative gastrocnemius), the main difference between S and C lines across different P_O2 seems to be associated with central factors in the O₂ cascade. Our results also show that individual variation in aerobic performance can be independent of P_O2; that is, individuals with high aerobic performances in normoxia consistently achieve higher V̇_O2max in hypoxia and hyperoxia (Table 4). In addition, selection history seems to be the major factor determining how individuals will perform at different P_O2.

Because mice attained higher V̇_O2max in hyperoxia regardless of selection history, our study shows that V̇_O2max in normoxia seems to be centrally limited in these lines. However, in a previous study (Rezende et al., 2005),we used cold exposure in combination with a He-O₂ atmosphere to demonstrate conclusively that lungs and heart can provide more O₂than necessary to attain V̇_O2max on the treadmill, a result that indicates that central limitations do not seem to be applicable for V̇_O2max during forced exercise. Taken together, these results suggest that whole-organism V̇_O2max is probably dependent on the interaction of several factors at many different levels in the O₂ cascade, as has been suggested in recent studies of metabolic regulation (e.g. Lindstedt and Conley, 2001; Darveau et al., 2002; Suarez and Darveau,2005). Nevertheless, the differential response to O₂availability between S and C lines suggests that subordinate traits have evolved somewhat independently and thus that S lines have not evolved in a strictly symmorphotic fashion.

We have also observed that mice with the mini-muscle phenotype achieved significantly higher V̇_O2max than normal only under hypoxic conditions, which indicates a genotype-by-environment interaction. This interesting result seems consistent with the idea that phenotypes favorable for locomotor endurance might also evolve as a correlated response to selection for hypoxia tolerance(Hochachka et al., 1998). Further, it would seem that a selection experiment designed to increase hypoxia tolerance could be a novel way to explore such hypotheses about correlated evolution of complex phenotypes (see Swallow and Garland, 2005 and references therein). Finally, the mini-muscle phenotype may prove an interesting model to elucidate genetic components underlying increased hypoxia tolerance and its evolution (Powell,2002). Although the reasons why the mini-muscle phenotype achieves higher V̇_O2maxonly in hypoxia are yet unknown, we hypothesize that higher cardiac outputs and muscle vascularization, as well as increased P_O2 gradients due to enhanced muscle O₂ extraction, are involved.

List of abbreviations
 
• V̇ _O2

  oxygen consumption (ml O₂ min^-1)

 
• V̇ _O2max

  maximum oxygen consumption observed during forced exercise (highest 1-min average)

 
• V̇ _CO2

  CO₂ production (ml CO₂ min^-1)

 
• V̇ _CO2max

  maximum CO₂ production during forced exercise (highest 1-min average)

 
• RER_V̇O2max

  respiratory exchange ratio(=V̇_CO2/V̇_O2)during V̇_O2max

 
• RER_V̇CO2max

  respiratory exchange ratio during V̇_CO2max

 
• P_O2

  oxygen availability (14, 21 and 30% O₂)

We thank L. Karpinski, J. Sinclair and several undergraduates for their help with the mouse colony, and J. Urrutia for constructing the treadmill. All animal procedures are in compliance with the UCR Institutional Animal Care and Use Committee and US laws. This study was supported by NSF IBN-0111604 (to K. A. Hammond and M.A.C.) and NSF IBN-0212567 (T.G.).