Table 1 presents the baseline physiological characteristics of the HRG and LRG. As can be observed, no differences in baseline demographics or physiological characteristics existed between LRG and HRG subjects prior to the training program 8. The combined training response was significantly greater in the HRG (p < 0.001). The reduction in submaximal heart rate was more substantial (p < 0.01) in the HRG (-25.9 ± 3.6 beats min^-1 (BPM)) vs. the LRG (-10.5 ± 3.5 BPM) during a 10 min fixed-workload, submaximal cycle. The increase in work performed during 15 min cycling after the six weeks of endurance exercise training was 60% greater (p < 0.05) in the HRG (+40.8 ± 4.3 kJ) than in the LRG (+24.9 ± 4.1 kJ). The increase in peak rate of oxygen consumption (peak VO₂) was four times greater (p < 0.001) in the HRG (+0.71 ± 0.1 L min^-1) than in the LRG (+0.17 ± 0.1 L min^-1). Importantly, there was no correlation between baseline physiological characteristics and the magnitude of improvement noted; this observation is consistent with those from previous studies 31.

Subjects in both groups underwent six weeks of endurance training. Gene expression levels were then measured 24 h after the last training session. Levels of VEGF gene expression were not significantly altered in either group (Fig. 1). Likewise, levels of VEGF receptor 1 (VEGFR1) mRNA did not significantly increase in the HRG and actually declined in the LRG (p < 0.05). However, VEGF receptor 2 (VEGFR2) mRNA expression increased by threefold in the HRG (p < 0.01), whereas it did not significantly change in the LRG. Similarly, mRNA for the VEGFR2 coreceptor Neuropilin-1 (NP-1) increased in the HRG (p < 0.001) but not in the LRG (Fig. 1). Expression of the hypoxia-inducible factor 1α (HIF) gene did not significantly change after endurance training. Regulation by endurance training of ANG-related genes was only observed in the HRG (Fig. 2). Levels of mRNA coding for ANG1, an agonist for the Tyrosine kinase with immunoglobulin-like and EGF-like domains 2 (TIE2) receptor, increased significantly in the HRG (p < 0.05), although ANG2 levels did not change significantly in either group. Likewise, transcription of tyrosine kinase with immunoglobulin-like and EGF-like domains 1 (TIE1; 3.4 ± 1.0-fold increase; p < 0.01) and TIE2 increased in the HRG only (p < 0.01; Fig. 3).

Using our microarray dataset 8, we selected ECM factors that (a) demonstrated evidence of being modulated by exercise training and (b) were relevant for ECM remodeling 323334. These genes included ones that encoded structural components of blood vessels (collagen type IIIα1) or known regulators of tissue angiogenesis (collagen type XVα1), factors known to influence ECM-derived growth factor activity (α2-macroglobulin (A2M) and thrombospondin-4 (THBS4)), transforming growth factor β2 (TGFB2; a potent regulator of tissue remodeling) and TGFB receptor II (TGFBR2). The level of fetal vascular collagen (i.e., collagen type IIIα1) gene expression was increased 14.7 ± 2.8-fold (p < 0.001) in the HRG and 8.5 ± 4.1 fold in the LRG (p < 0.001). The level of collagen type XVα1 expression increased in both groups: the HRG demonstrated a 7.6 ± 3.6-fold increase (p < 0.001) while the LRG demonstrated a 2.7 ± 0.5-fold increase (p < 0.01) in expression of the gene (Fig. 3). In contrast, while both A2M and THBS4 were significantly upregulated in the HRG (threefold (p < 0.01) and tenfold (p < 0.001) increases, respectively), in the LRG the THBS4 response was more modest (2.5-fold; p < 0.01) and A2M mRNA levels did not significantly change. TGFB2 was significantly upregulated in the HRG (p < 0.01), whereas, if anything, it tended to be downregulated in the LRG (Fig. 3). The level of TGFBR2 expression was unchanged in the LRG, but threefold increased in the HRG (p < 0.001; Fig. 3). Finally, three genes unrelated to ECM biology but previously described as being modulated by exercise 8, interleukin 17D, Rho-GTPase-activating-protein 1 and myristoylated alanine-rich protein kinase C substrate, did not significantly vary in their expression between HRG and LRG (data not shown).

We still have an incomplete understanding of the endogenous processes that regulate physiological adaptation to aerobic exercise. Vascular growth factors not only regulate tissue blood vessel density, but also enhances the expression of proteins that regulate oxygen levels in tissue 37. Hence, it is naïve to think about the role of such growth factors only in terms of regulating tissue capillary density. Altered expression of HIF responsive genes (e.g. VEGF or VEGFR1) typically reflects the posttranslational stabilization of HIF1α protein 38, and consistent with this dogma, HIF mRNA was not significantly upregulated. VEGFR2, considered the major mediator of VEGF-A-related angiogenesis, was significantly upregulated in the HRG only. Furthermore, NP-1, a facilitator of VEGF₁₆₅ action at VEGFR2, was markedly elevated in the HRG and unchanged in the LRG (Fig. 1). Upregulation of the VEGFR2 coreceptor transcript provides some evidence for greater VEGF activity in the HRG, as enhanced NP-1 gene expression can be mediated by VEGF signaling 3940. The significance of the downregulation of VEGFR1 in the LRG is unclear, but plausibly reflects an (unsuccessful) compensation for the general lack of VEGFR2/NP-1 gene activation in the LRG, as studies indicate VEGFR1 may oppose VEGF signaling via VEGFR2 in some situations 41.

The present study represents the first characterization of expression levels of the human ANG gene family in response to endurance-exercise training. Recently, the Terjung laboratory examined the impact of exercise on the angiopoietin system in rodents 27. Alterations in TIE2, ANG1 and ANG2 transcript expression were profiled 2 h post exercise in various muscle groups taken from Sprague-Dawley rats after 1 to 24 days of intense aerobic running 27. In rodents, activation of TIE2 and ANG2 was, broadly speaking, rather similar across each muscle tissue, especially when one considers that recruitment patterns would differ between the various muscle groups studied. The ANG1 response, however, differed both across muscle groups and when compared with our human data. In contrast with the upregulation of ANG1 seen in humans in the HRG (Fig. 2), in the rat ANG1 expression was slightly downregulated in oxidative muscle groups, while 'fast' gastrocnemius demonstrated only a modest increase in ANG1 expression 27. It has recently been established that either pre- 30 or concurrent administration 29 of ANG1 synergizes with VEGF to promote hindlimb angiogenesis. Thus it makes sense that effective aerobic training might result in stable increases in ANG1 expression (as we have found). Therefore, differences between our human data and the rodent study by Lloyd et al 27 perhaps reflect differing muscle sampling times. Future human studies should ideally use multiple time points post exercise to clarify these issues. However, care should be taken to verify that the subjects studied are able to demonstrate a measurable aerobic training response otherwise such results may be unreliable.

It has been hypothesized that changes in the ratio between levels of ANG1 and ANG2 is of physiological importance 2742 by contributing to the stabilization of the primary endothelial structures 43. While ANG1's ability to antagonize ANG2 at the TIE2 receptor may be cell-type specific 424445 and has yet to be proven to be important in vivo, it is interesting to note that ANG1 and ANG2 appear to have identical binding affinities for the TIE2 receptor, whereas a threefold molar excess of ANG2 is required to antagonize ANG1 activity 45. If the gene expression patterns observed in the present study translate to greater Tie2 receptor signaling, then our data supports the idea that ANG1 cooperates with other growth factors in vivo to regulate and promote functional angiogenesis. This hypothesis clearly requires further investigation. ANG1 and TGFB signaling facilitate the maturation of VEGF-stimulated collateral vessel growth in adult tissue 293046. Notably, TGFB2 and TGFBR2 were substantially upregulated only in the HRG. This observation again suggests that the ECM-related gene response in the HRG may have allowed for greater tissue remodeling, which would have contributed to the fourfold greater increase in aerobic capacity.

To further examine the idea that the pattern of transcript expression in the HRG contributed to the enhanced aerobic adaptation, we profiled a second set of genes chosen from a gene-array study 8. For example, upregulation of potent growth factors should be accompanied by upregulation of endogenous regulators, so that physiological control could be maintained. A2M is an ECM protein known to bind and regulate growth factor activity 4748. A2M was substantially upregulated in the HRG only (Fig. 3) suggesting that there was more active growth factor signaling within the ECM of the HRG. Thrombospondins also regulate ECM growth factors, including TGFB activity. In addition, a loss of function polymorphism in THBS4 is strongly associated with premature coronary artery disease 49. In the present study a more substantial increase in THBS4 mRNA expression was noted in the HRG (Fig. 3). As dysfunction of THBS4 and lack of cardiorespiratory fitness are both risk factors for cardiovascular disease, it is plausible that THBS4 plays a role in the cardioprotective effects of exercise. Further analysis is required to establish if such a relationship exists within a larger human population.

The study was approved by the ethics committee of the Karolinska Institutet, Stockholm, Sweden, and informed consent was obtained from each of the volunteers. Subjects abstained from strenuous exercise during the three weeks prior to obtaining pretraining muscle biopsies (taken from the vastus lateralis). Twenty-four subjects trained under full supervision on a cycle ergometer four times a week (45 min) at 75% of their pretraining peak VO₂ for six weeks. Posttraining biopsies were taken 24 h after the last training session. Physiological measurements and muscle biopsies were performed as previously described 81650. All physiological parameters were derived from a minimum of two assessments, taken on separate days. Peak VO₂ was determined using a cycle ergometer (Rodby, Sweden). An incremental protocol was combined with continuous analysis of respiratory gases using the Sensormedic ventilator (Sensormedic Co., USA). At peak VO₂ the respiratory exchange ratio and heart rate exceeded 1.1 and 190 BPM, respectively, on all occasions. Total amount of work done in 15 min of cycling was determined using a self-paced protocol (Lode, Netherlands, test-re-test variability <5%). Submaximal physiological parameters were determined during two separate 10 min constant-load, submaximal cycling sessions (carried out at 75% of pretraining peak VO₂).

Two groups were selected from a larger training cohort based on the magnitude of their training response (Table 1). The two groups represented the eight subjects that demonstrated the largest improvement in physiological capacity (HRG) and the eight subjects that demonstrated minimal changes in physiological capacity (LRG) after following an identical supervised training program. The average training-induced change in peak VO₂ in the LRG equates to the lower ~10% of responses observed in the HERITAGE study 6, whereas the response in the HRG represents the top ~10% of responses. This information helps to establish a 'population perspective' on the responses observed in the present study. Subjects were assigned to each group after being ranked on the basis of the sum of changes in three main physiological parameters: (a) percent improvement in peak aerobic capacity, (b) percent reduction in submaximal heart rate during 10 min fixed-workload, submaximal cycling and (c) percent improvement in work done during a 15 min maximal cycling test. A combination of physiological markers of adaptation was used to ensure that no spurious individual physiological measurement would compromise the overall categorization of a subject 8. Peak aerobic capacity, decreased exercise-induced-increases in heart rate and aerobic performance are all valid and accepted responses to aerobic exercise training. No molecular or biochemical analysis was carried out until the subjects were assigned to their particular group. This type of novel analysis strategy has previously been used by our group and also, more recently, by Bouchard et al 89

Total RNA was prepared using the TRIzol method (Invitrogen, USA) and quantified using a spectrophotometer. Two μg of RNA was reverse transcribed by Superscript reverse transcriptase (Life Technologies, Sweden) using random hexamer primers (Roche Diagnostics GmbH, Mannheim, Germany) in a total volume of 20 μl. Detection of mRNA was performed using a ABI-PRISM^® 7700 Sequence Detector (Perkin-Elmer Applied Biosystems Inc, Foster City, CA, USA). All reactions were performed in 96-well MicroAmp Optical plates (Perkin-Elmer Applied Biosystems Inc.). Amplification aliquots contained 5 μl of the sample cDNA, the TaqMan Universal PCR master mix (Perkin-Elmer Applied Biosystems Inc.) and an optimized concentration of each primer and probe, prepared according to the manufacturer's recommendation, in a final volume of 25 μl. 18S rRNA was selected as an endogenous control to correct for potential variations in RNA loading into the cDNA synthesis reaction. The 18S rRNA control was run in triplicate in separate wells (using 1:2000 dilution of the original cDNA). Thermal cycling conditions were, initially, 2 min at 50°C followed by 10 min at 95°C, and then, subsequently, 45 cycles of 15 s at 95°C and 1 min at 65°C.

Oligonucleotide primers and TaqMan probes were designed using Primer Express version 1.5 (Perkin-Elmer Applied Biosystems Inc.) and synthesized by Cybergene (Stockholm, Sweden) or ordered as a 'gene assay by demand' product (Perkin-Elmer Applied Biosystems Inc). The sequences or 'gene assay by demand' numbers can be provided on request. The probes were designed to cover exon-exon boundaries to avoid amplification of genomic DNA. As there are no predetermined low-abundance "house-keeping" genes for this experimental paradigm, the ΔΔCt method 51 was used to calculate relative changes in mRNA abundance. The threshold cycle (Ct) for 18S was subtracted from the Ct for the target gene to adjust for variations in mRNA/cDNA generation efficacy. This was carried out for both pre- and posttraining samples. The preexercise value reflects baseline gene expression levels and was subtracted from the postexercise value to calculate the increase or decrease in mRNA abundance.

A two-way mixed model ANOVA (GraphPad Prism 4.0) was used to establish whether a significant interaction between 'group' and 'extracellular gene responses' occurred (p < 0.0001) and to confirm appropriate baseline subject matching (p < 0.0001). Bonferoni post-hoc testing established which individual gene responses were significant for each subgroup, and this analysis was used for the basis of the discussion. For comparison between HRG and LRG training response data (e.g., peak VO₂), a two-tailed unpaired t-test was utilized. Significance was accepted at the 5 % level. Data are mean ± SE.