Intermuscular variation in quadriceps architecture (absolute differences)
In order to produce forces with broad magnitude, range and velocity characteristics, it has been suggested that muscles within synergistic groups tend to vary in their architecture (Lieber & Fridén, 2000). Our examination suggests that this is only partly true for the human quadriceps muscles. The difference index calculated to provide quantitative assessment of intermuscular variability revealed a close similarity between the mean architecture of the three superficial quadriceps muscles (i.e. VL, VM and RF; δ2−1 < 0.36), despite RF being bi-articular and VL and VM being mono-articular. Given that muscles within the human quadriceps femoris act via a common (patellar) tendon, this similarity in mean architecture is indicative of a similarity in their force-generating potential during uni-joint knee extension (where RF functions as a mono-articular knee extensor), although their angle of attachment to the tendon would also influence their contribution to knee function. Nonetheless, the deeper VI is architecturally dissimilar to the superficial muscles, indicating that it has a different force-generating potential, and possibly a different role during knee extension. A clear picture of its role is difficult to ascribe because its architecture varies significantly. The anterior portion is characterized by long fibres attaching at small angles to the aponeurosis, whereas the lateral portion has shorter fibres attaching at larger angles (δ2−1 = 1.97); the medial portion was not examined.
This intermuscular variation suggests that the superficial and deep quadriceps muscles have different roles, and could be recruited uniquely based on their force-generation properties. Such recruitment variability has been previously shown. In the quadriceps femoris, Zhang et al. (2003) found that the relative contribution of each muscle to the total isometric knee extensor moment changed as contraction force increased, such that the contribution of VI decreased whereas the contribution of VL, VM and RF increased (the increase in RF approached statistical significance, P = 0.06). Similar findings have been reported for the biceps brachii, which also consists of multiple muscles attaching via a common tendon (Brown et al. 1993). Nonetheless, other studies employing surface EMG analyses have also shown muscle-specific changes in quadriceps muscle activation as force levels are varied (Ebersole et al. 1999; Pincivero & Coelho, 2000; Rabita et al. 2000; Pincivero et al. 2002, 2004b; Cramer et al. 2004), knee joint angle, and therefore whole quadriceps femoris muscle length, is changed (Ebersole et al. 1999; Signorile et al. 1995), velocity is increased (Cramer et al. 2004) or contraction mode is changed from concentric or isometric to eccentric (Narici et al. 1996; Pincivero et al. 2006). These findings are not predicted by a model that assumes a similarity in architecture of the superficial quadriceps muscles. However, these variations might be explained by differences in intramuscular architecture, as discussed below (‘Homogeneity of within-muscle architecture’). Movement-specific muscle activation strategies will have to be examined further as a relationship between muscle architecture and recruitment has yet to be demonstrated. Notwithstanding, there is some support for hypothesis 1, that the architecture of each muscle within the quadriceps should be different, for the superficial vs. deep quadriceps muscles; however, there seems to be some uniformity among the superficial muscles.
The architectural differences between muscles would also affect the efficiency of force transfer to the tendon, given that forces are transferred laterally between muscles (e.g. Huijing, 2003; Maas & Huijing, 2005). For example, the shorter RF fascicles attaching at larger angles (12.5°, 112.3 mm length at 45° knee angle) would be expected to generate high relative peak forces over shorter length ranges, which is commensurate with reports that it works quasi-isometrically during many human movements (Gregoire et al. 1984; Ingen Schenau et al. 1987; Jacobs & Ingen Schenau, 1992). This can be compared with the immediately adjacent, architecturally dissimilar VIant (6.5°, 162.1 mm at 45° knee angle), which is suited to generating lower peak forces over greater length ranges. Even when fibre rotation is accounted for, which would allow RF fascicles to shorten less relative to the whole muscle shortening, there is still a significant force–length discrepancy between the muscles. As shown in Fig. 4, a similar relative shortening of RF and VI (anterior) fascicles would result in a different tendon or aponeurosis displacement (10.0 mm vs. 14.2 mm in the example). The discrepancy would result in a shear stress between their aponeuroses, and is likely to affect force transfer efficiency.
Figure 4. Effect of intermuscular fascicle length and angle differences on tendon/aponeurosis excursion. When rectus femoris (RF) fascicles shorten (from FLRF1 to FLRF2, where FL is the fascicle length) and rotate (θRF1 to θRF2, where θ is the fascicle angle) in this example, the aponeurosis or tendon displaces 10 mm. This represents a fascicle shortening of 9.7 mm, or 8.7%. If the vastus intermedius (VI; anterior region) fascicle shortens by the same proportion, 14.0 mm, or 8.7%, and rotates, the aponeurosis or tendon displaces 14.2 mm, which is a difference (Diff.) of 4.2 mm, or 42%. Thus, even when fascicle rotation is accounted for, the differing architectures of RF and VI result in differential displacement of the tendon or aponeurosis. Either RF and VI must be activated uniquely, or there will be a shear force in the tendon or aponeurosis. In this example, it was assumed that the muscle thickness would remain constant given that RF and VI apply opposing forces of similar magnitude on the tendon/aponeurosis.
Download figure to PowerPoint
Alternatively, a lesser recruitment of VI as contraction intensity increases (see Zhang et al. 2003) would allow matching of muscle shortening. Because VIant contains longer fibres, and presumably a greater number of sarcomeres in series, a similar absolute shortening of RF and VI fascicles would result in a lesser shortening of VI sarcomeres. This would improve force production via optimization of force–length and force–velocity relationships and might allow more efficient force production. As discussed above, variations in the recruitment of individual muscles has been demonstrated previously, but the hypothesis that the strategy of unique muscle activation would improve force transfer efficiency has yet to be examined. Data from near-infrared spectroscopy experiments has shown that the increase in O2 utilization, and therefore energy consumption, by the superficial quadriceps muscles with increasing force levels is similar between muscles (de Ruiter et al. 2005). These data are consistent with the hypothesis that the unique activation of muscles optimizes efficiency under different loading conditions. Furthermore, strength training of the quadriceps has not only been shown to alter the proportional activation of the quadriceps muscles (Pincivero et al. 2004a) but has been shown to reduce the EMG measured from superficial quadriceps muscles at prescribed submaximal force levels (i.e decreased EMG–force ratio). This latter change is also indicative of a greater efficiency of force production (Narici et al. 1996; Rabita et al. 2000). The contention that unique muscle activation would be a useful optimization strategy also suggests that the EMG–force ratios of individual muscles, and their adaptation with movement training, would be specific to the person as architecture varies considerably interindividually. This contention has been given some credence by Rabita et al. (2000), who showed that alterations in the EMG–force relationship varied significantly between muscles and also between individuals. Thus, although direct evidence that humans adopt a unique activation strategy in order to minimize shear forces and improve force transfer efficiency in the human quadriceps is not available, published data are strongly supportive of the concept.
Importantly, many muscle models assume that architectural variations have little impact on the active length range of muscles working in synergistic groups (see Van den Bogert et al. 1998), largely because it is supposed that connective tissues would be overstretched/injured if significant discrepancies existed. Based on our data, this assumption might need to be revised; significant intermuscular architectural differences might explain some of the difficulties in accurately modelling complex muscle groups. Thus, the effects of architectural dissimilarity in muscles functioning as a synergistic group are complex, and more research is required to understand activation strategies used to optimize muscle function. Regardless, the quadriceps femoris probably cannot be accurately modelled as a muscle group that provides a force exactly equal to the sum of each of its constituents.
Homogeneity of within-muscle architecture (absolute differences)
Although mean architecture measures were relatively consistent between muscles, except for VI, our data also show that architecture does vary significantly within them. For example, fascicle angle increased proximo-distally in VM, decreased proximo-distally in RF and was greater centrally in VL. Furthermore, fascicle angle was significantly different in anterior and lateral regions of VI, although it changed little along the length of the muscle. These results are consistent with previous reports of intramuscular architectural variation (Scott et al. 1993; Blazevich et al. 2003) and provide little support for our second hypothesis, that the intramuscular architecture should be relatively homogeneous. They are also suggestive of the possibility that different regions of the muscles have different functional roles, and perhaps are activated uniquely during knee extension. Although there are no data describing region-specific variations in activation of the individual quadriceps muscles, different regions of other muscles such as the long head of biceps brachii are differentially activated as force patterns are altered (ter Haar Romeny et al. 1984). A similar examination of the quadriceps muscles might reveal region-specific activation patterns.
Intramuscular architectural variation is probably required in pennate muscles in order to achieve mechanical stability (Van Leeuwen & Spoor, 1992), although such variation might further influence muscle function and force transfer between muscles. For the same relative shortening of fascicles in different regions of the muscles, aponeurosis displacement will be different. Most planimetric models of muscle assume a constant fascicle angle along the length of a muscle, which might not represent the muscles appropriately. More complex muscle models that allow fascicle angle and length parameters to vary (Woittiez et al. 1984; Otten, 1988; Savelberg & Schamhardt, 1995) will probably be more accurate. To estimate the extent to which models might be affected when architecture is assumed to be constant, we constructed a planimetric model of a unipennate muscle containing three segments, or parallelepipeds (Savelberg & Schamhardt, 1995), as shown in Fig. 5. It was assumed that each segment contributed to the tendon force in proportion to its estimated physiological cross-section (muscle thickness2/fascicle length), and the total muscle force was equal to the sum of forces provided by each segment (see Appendix A). The model was run with three different inputs: (1) where each segment was identical in its architecture, with mean VL (middle) data being assumed to run throughout the muscle; (2) where each segment varied in its architecture, with mean VL proximal, middle and distal data being used as input for the three segments; and (3) where each segment varied, with VL (middle) data being used as input for the middle segment but mean − 1 SD data being used from distal and proximal VL as input for segments 1 and 3 (see Fig. 5). The results of the model show little difference in the force–length curves when the muscle is assumed to have a constant architecture compared with a varying architecture (Fig. 5A vs. Fig. 5B) when the architecture varies within the orders of magnitude reported for VL in the present study. However, when more significant architectural variation is input into the model (Fig. 5C), the force–velocity relationship changes far more dramatically. Muscles that vary considerably in their architecture, such as VM, will probably not to be accurately modelled when intramuscular architecture is assumed not to vary, so consideration must be given as to whether uniform muscle models are appropriate for use.
Figure 5. A planimetric muscle model was constructed, according to Appendix A. The muscle consisted of three segments which were either identical in their architecture (non-varying architecture, A), varied according to normal data for the VL collected in the present study (varying architecture, B), or varied more considerably in the proximal and distal segments than the measurements in the present study (i.e. mean minus 1 SD) (varying architecture, C). Muscle thickness (MT) and fascicle angle (θ) parameters used in the model are shown (parameters correspond to muscle optimum length). As shown in the graph of muscle force vs. muscle length (D), modelling the muscle as three distinct segments makes a relatively small difference to the output of the model when architecture varies only moderately (A vs. B). However, when the muscle's architecture varies considerably (A vs. C), the ability for a uniform muscle model to predict muscle force is significantly reduced.
Download figure to PowerPoint
The functional implications of intramuscular variations in architecture are unclear; however, we posit two hypotheses (additional to the requirement for mechanical stability; Van Leeuwen & Spoor, 1992). First, the muscles might be intended to have both ‘force-generating’ and ‘force-transfer’ regions. The greater fascicle angle and subsequently larger physiological cross-section in one region might perform the function of generating high muscle forces. A second, force-transfer, region might be designed to have fibres in line with the tendon to allow greater efficiency of force transfer. The use of a transfer section consisting of muscle, rather than tendon, would permit greater control over muscle–tendon stiffness, and subsequently over movement control. The second hypothesis is that fascicles within muscles tend to align with the joint's centre of rotation, which would improve the efficiency of force transfer from muscle to bone. The present data cannot confirm this hypothesis because the angle of the aponeurosis relative to the tendon was not measured, so pennation angle could not be determined. However, our findings that fascicle angle generally increased in VM proximo-distally, but not in RF and VI, and that the lateral portion of VI had greater fascicle angles, tend to support the hypothesis, whereas the finding that fascicle angle decreased from middle to distal regions in VL is not supportive.
Intramuscular architectural variation should also affect the hypertrophic response to, and therefore the function of, different regions of the muscle. The variation in fascicle angle (and length) along these muscles would probably affect fascicle strain in response to an imposed stress (Maganaris & Paul, 2000; Muramatsu et al. 2002; Finni et al. 2003a). Given the assumption that the hypertrophic response of muscle is somewhat related to fibre strain (e.g. Goldspink, 1978; Vandenburgh & Kaufman, 1979; Gregory et al. 1986; Wong & Booth, 1990; Hather et al. 1991; Goldspink et al. 1992, 1995), one would predict region-specific muscle hypertrophy in response to muscle loading. Published data are consistent with this hypothesis (Housh et al. 1992; Narici et al. 1996; Häkkinen et al. 2001). Interestingly, both Häkkinen et al. (2001) and Narici et al. (1996) reported greater proportional hypertrophy proximally in VM and distally in VL; our data show that in these regions the fibre angles of both muscles are relatively smaller. Unfortunately, there are no data showing a direct link between local muscle architecture and the hypertrophic response. A direct conclusion is also made more difficult in view of the findings of Housh et al. (1992) whose subjects increased at the mid-level of VM and VL only. Furthermore, there is a lack of agreement as to the predominant sites of hypertrophy in VI and RF. It is possible that adaptations are exercise-specific, and might also be influenced by the activity profiles of the subjects used. Thus, although the hypothesis that intramuscular architectural variation should result in region-specific hypertrophic adaptations to exercise training has some support, a direct link between hypertrophy and pre-existing architecture has not yet been made.
Homogeneity of within-muscle architecture (relative differences)
Despite architecture varying considerably along the length of three of the quadriceps quartet, there was a constant intersubject variability of regional variation in VL, VM and RF. For example, subjects who had a greater proximal fascicle angle in a given muscle, relative to the other subjects, also tended to have greater fascicle angles at middle and distal sites in that muscle. This was not necessarily the case for VI. In the anterior portion, muscle thickness between sites was highly correlated (i.e. subjects with greater or lesser thickness at one region also had greater or lesser thickness at another region), but fascicle angle was only correlated between middle and distal regions. For the lateral portion, muscle thickness between regions was not correlated, and fascicle angle was only correlated between proximal and middle regions. Thus, while intersubject differences in one region of a superficial muscle were similar to those in other regions, and therefore in broad terms relative architecture in one region is a reasonable indication of relative architecture in another, there was less evidence of this for VI. Because our second hypothesis was that muscle architecture should be uniform throughout a muscle, it stands also that we would expect a subject's relative architecture (i.e. compared to the rest of the sample of subjects) at one region of a muscle to be indicative of its relative architecture at another region in all muscles within the quadriceps femoris. Thus, our hypothesis was proved to be incorrect.
Intermuscular variation in quadriceps architecture (relative differences)
Relative intermuscular variation was considerably greater than that within a muscle. For example, subjects who had a greater mean fascicle angle of VM also tended to have greater angles in VL and RF, and this was also true for RF, where greater fascicle angles were indicative of greater angles in VL and VM. However, this was not the case for VL and VI (both anterior and lateral regions), where relative architecture was not predictive of architecture in other muscles. Thus, although there is some indication that the architecture in one superficial muscle is reflective of that in another, this is not completely true for VL and does not hold true for the deep VI. These data are not supportive of hypothesis 3, that the relative architecture of one muscle should be similar to that of another. Therefore, to gain an overall impression of whole quadriceps architecture a description of more than one muscle is required.
Contribution of individual muscle architectures to whole quadriceps architecture
We also sought a global measure of architecture for the quadriceps femoris. The low mean intermuscular architecture variability in superficial quadriceps muscles was suggestive that this might be possible, although these muscles differed in a region-specific manner. We used four methods for estimating whole quadriceps architecture, including (1) allowing the mean architecture of each muscle an equal weighting (equal ratio), (2) weighting by relative muscle volumes (volume), (3) weighting by relative muscle physiological cross-sections (PCSA) and (4) weighting by estimated contribution to knee extensor torque as reported by Zhang et al. (2003; estimated contribution). There were no significant differences in muscle thickness, fascicle angle or fascicle length values found for the whole quadriceps between the methods (R2 > 0.92), but there were some differences in the level of contribution of each of the muscles. Most notably, whereas the equal ratio, PCSA and estimated contribution methods gave statistically similar results, estimation by volume resulted in VL thickness and fascicle length being more highly correlated with whole quadriceps thickness, and RF thickness being less highly correlated with whole quadriceps thickness. Furthermore, a statistically significant gender difference in muscle thickness (men > women) was found using the equal ratio, PCSA and estimated contribution methods, but not using the volume method. Given the above, we suggest caution in weighting architectural values by volume when estimating whole quadriceps architecture.
Prediction of whole quadriceps architecture from individual muscle architecture
An important benefit of estimating whole quadriceps architecture is that it allows the identification of individual muscles whose characteristics might be representative of the muscle group. An understanding of this would preclude the necessity to measure architecture of all four muscles in order to characterize an individual's whole quadriceps architecture. Unfortunately, according to the regression analysis (Table 6), estimation of whole quadriceps architecture requires the inclusion of data of up to four muscles: VL, VM, RF and VIant. This is not supportive of hypothesis 4, that the relative architecture of individual muscles should be indicative of relative whole quadriceps architecture, i.e. compared with other subjects, a subject's rank by magnitude for an architectural parameter in one muscle was not the same as their rank for the parameter in another. Most typically, VM architecture was the best predictor of quadriceps architecture, as evidenced by its primary inclusion in the estimation of fascicle angle and muscle thickness for almost all methods of estimating whole quadriceps thickness and fascicle angle (see Table 6). Because VM fascicle length measures were unreliable owing to the excessive curvature of the fascicles in some subjects, they could not be used to predict whole quadriceps fascicle length. In this instance, both VL and VIant were the primary predictors. In general, VL and VIant were most frequent secondary predictors. VIlat was excluded in all instances and is therefore considered a poor estimate of whole quadriceps architecture. Thus, VM architecture was generally the best predictor of whole quadriceps architecture, although the inclusion of at least VL and/or VIant is needed for more accurate estimates. Although these muscles might allow a ‘representative snapshot’ of the whole quadriceps to be drawn, it is not known whether gross changes in quadriceps architecture that result from physical training or detraining can be quantified by measurement of only two of the quadriceps muscles. This needs to be determined from studies specifically examining architectural adaptation.
Validity of whole quadriceps estimates: interaction between muscle thickness, fascicle angle and fascicle length
On the whole, our estimates of whole quadriceps architecture appear to be a valid representation as they predict two important relationships that are normally found in individual muscles. First, a gender difference in whole quadriceps thickness was obtained for all methods except when muscles were apportioned by their relative volumes. This finding of larger muscles in men than women is consistently reported in the literature (Abe et al. 1998; Kubo et al. 2003). It is interesting to note that fascicle angles and lengths did not differ significantly between sexes. While relatively few studies have reported sex variation, Chow et al. (2000) showed differences in muscle size, fibre length and fibre angle between men and women in the gastrocnemius and soleus muscles in non-athletic, younger subjects (Chow et al. 2000). Abe et al. (1998) observed that muscle thickness and fibre angle differences in triceps brachii, vastus lateralis and gastrocnemius muscles were also different between height- and weight-matched, resistance-trained men and women. It is not clear whether differences in bodyweight (not reported by Chow et al. 2000) or fat-free body mass (men > women; Abe et al. 1998), or muscle size (Abe et al. 1998; Chow et al. 2000), might at least partly explain these differences given that greater muscle size is generally associated with either, or both, increased fascicle angles or lengths. Nonetheless, other reports suggest there is no difference between male and female trained sprinters (Abe et al. 2001), Olympic soccer, gymnastics and judo athletes (Ichinose et al. 1998). Given that there are sex-dependent differences in muscle force production and rates of injury, more effort should perhaps be directed at understanding sex-dependent architectural differences.
Our finding of a lack of sex difference in the quadriceps also implies that muscle activation strategies, and their changes with exercise training, should also be similar. This is consistent with published data showing that alterations in the EMG–force relationship of the muscles are the same in men and women as movement velocity (Cramer et al. 2004) and contraction mode (Pincivero et al. 2006) are varied. However, it is not consistent with the finding that the proportional contribution of VL increased more in men, and the contribution of VM and RF decreased in women, as knee extension contraction force increased (Pincivero et al. 2002). Because muscle architectural measurements were not performed in these studies, it is not possible to ascribe their inconsistent findings to differences in muscle architecture, although other factors such as differing exercise patterns between the men and women need also to be assessed. To our knowledge, no research has examined training-related changes in quadriceps activation between men and women. The idea that alterations in synergist activation patterns might be the same in men and women has particular implications for clinicians and therapists whose goal is to optimize movement efficiency rapidly in patients after periods of disuse.
A second indicator that our whole quadriceps architecture estimates appear to provide a valid representation of the muscle group is that the data show a significant positive correlation (r > 0.65, P < 0.001) between muscle thickness and fascicle angle regardless of the method used to estimate whole quadriceps architecture, and a weaker negative correlation between fascicle angle and fascicle length (r > 0.45, P < 0.05) (see Fig. 6), and are thus in agreement with others (Abe et al. 1998; Kumagai et al. 2000; Brechue & Abe, 2002; Kanehisa et al. 2003). In general, the highest correlations were achieved using the ‘estimated contribution’ method of estimating whole quadriceps architecture, which might give some speculative validation to the results of Zhang et al. (2003). In some respects, these whole quadriceps data should be treated with some caution because, for example, fascicle length was positively correlated with muscle thickness in RF, suggesting that important architectural characteristics of individual muscles might be missed if only whole quadriceps values are used for analysis.
Figure 6. Relationships between whole quadriceps muscle thickness and fascicle angle (top graph: a), and fascicle angle and fascicle length (bottom graph: b). Whole quadriceps architecture in these examples was calculated by the equal ratio method (see Methods), which was associated with the lowest correlations between the variables when compared with PCSA, volume and estimated contribution methods. Muscle thickness and fascicle angle were positively correlated, while fascicle angle and fascicle length were negatively correlated. Magnitudes are presented as z-scores.
Download figure to PowerPoint