Central rib and the nutritive value of leaves in forage grasses

In grasses, leaf expansion and central rib growth occur in a non-proportional manner, with potential implications to the nutritive value of leaves. The objective of this study was to evaluate the relationship among blade length, percentage of central rib, anatomical characteristics and the nutritive value along the length of leaf blades of different sizes and hierarchical order of insertion on the tiller axis of Napier elephant grass (Pennisetum purpureum Schum. cv. Napier). Two experiments were carried out with isolated growing plants during the summer of 2017 (January to March). Central rib mass increased linearly with the increase in leaf blade mass and its percentage relative to blade mass decreased from the base to the tip of the leaf. There were no significant variations in anatomical characteristics along the length of leaf blades when central rib was not taken into account. The central rib showed negative relationship with nutritive value. The apical portions of long leaves showed similar digestibility to short leaves. The multivariate analysis of Cluster and Principal Components grouped the response variables according to leaf hierarchical order, final blade length and percentage of structural tissues, highlighting the relationship between leaf size, structural tissues and nutritive value.

In forage grasses, leaves are usually the morphological component with highest nutritive and feeding value. Their development start as primordia located on opposite sides of tiller apical meristems 1 , with blade length varying according to their order of appearance on the tiller axis (hierarchic order of insertion) and greatest values normally recorded on leaves of intermediate levels of insertion 2,3 . Morphologically, the leaf blade is normally linear or lanceolate shaped, with characteristic parallel ribs 4 , the main one being the central rib, which provides additional structural support to the leaf 5 . Leaf growth is developmentally coordinated with the growth of the major veins, given that the bulk of leaf expansion results from leaf length. For that reason, having a structurally reinforced central rib (i.e. more diameter growth of the midvein with greater elongation of the blade) is biomechanically necessary and reduces the increasing pressure required to conduct water with increasing path length 6 .
Anatomically, the leaf blade is comprised of vascular tissue (vascular bundles), represented by xylem and phloem cells; ground or support tissue, represented by parenchyma, collenchyma and sclerenchyma which in grasses is frequently associated with vascular tissue; and assimilatory tissue, represented by mesophyll cells 7 . These tissues are covered on both sides of the blade (abaxial and adaxial) by the epidermis that, in turn, is covered externally by the cuticle. At the same stage of development, leaves show an anatomical tissue gradient depending on their level of insertion on the tiller according to which blades and sheaths from leaves ranked higher on the tiller axis (closer to the apical meristem) have greater proportion of sclerenchyma and vascular tissue relative to those ranked lower (closer to the base of the tiller) 8 .
Leaf blades of higher insertion levels are usually longer providing that there is no elongation of the the apical meristem (internode elongation) [9][10][11] . As a result, they may have greater proportion of sclerenchyma and vascular tissue due to the need for structural support to maintain its erect architecture for intercepting the incident light 12 . The sclerenchyma along with the xylem cells are important to provide support to the erect growth of the leaves, particularly in warm and dry environments 13,14 . These tissues are formed by cells with thickened secondary walls and are the main determinants of reduced herbage quality 15,16 . In this context, as the central rib usually contains large proportions of vascular tissue and sclerenchyma, it has digestibility considerably smaller than the blade 17 . However, both shape and distribution of central rib mass vary along the leaf length, since central rib as well as second order parallel veins taper from the base to the tip during leaf growth, resulting in leaf tips narrower than leaf base; the lanceolate shape 4, 6, 18 .  Leaf anatomy. There was a slight variation in the percentage of anatomical tissues on the leaf blades as their level of insertion on the tiller axis increased. The percentage of both abaxial and adaxial epidermis slightly decreased and that of vascular tissue and sclerenchyma slightly increased as leaf size increased from phytomers 8 to 16 (Table 2). On the other hand, anatomical characteristics were quite stable along the blades (from the basal to the tip portion) for all the studied phytomers (Table 3). Only a slight variation in anatomical characteristics was observed along the length of the leaves as described above for leaf morphology. For the percentage of mesophyll, a common pattern of variation was detected for all phytomers, according to which greater and smaller percentages were recorded on the basal and mid-apical/ apical portions of the leaves, with variations of 45-52; 43-48; 42-47; 41-46 and 40-44% of the total area of the segments 1, 2, 3, 4 and 5, respectively (Fig. 3a). The same pattern was observed for sclerenchyma percentage, with variations of 0.9-2.0; 0.4-1.5; 0.3-1.3; 0.2-1.0 and 0.1-1.0% of the total area of the segments 1, 2, 3, 4 and 5, respectively (Fig. 3b). For the parenchymatic sheath of the vascular bundles and vascular tissue, greater percentage was observed on the mid-basal and middle segments (2 and 3), with variations from 18 to 26% of the total area of the segments (Fig. 3c). For the adaxial epidermis, greater percentage was observed on the apical portion (segment 5), values ranging from 20 to 25% of the total area of the segments (Fig. 3d). For the abaxial epidermis greater percentage was observed on the mid-apical/apical portions of the leaves, values ranging from 12 to 17% of the total area of the segments (Fig. 3e). The response patterns of anatomical composition relative to plant age (harvest of leaves when the 16th phytomer completed expansion and exposed the ligule) were similar to those described for individual phytomers soon after their complete expansion. The variations were: mesophyll-47-52; 44-49; 43-47; 42-46 and 41-44% of the total area of the segments (Fig. 4a); sclerenchyma-0.7-2.0; 0.5-1.8; 0.4-1.4; 0.2-1.2 and 0-0.6% of the total area of the segments (Fig. 4b), both for segments 1, 2, 3, 4 and 5, respectively; parenchymatic sheath of the vascular bundles and vascular tissue-17-25% of the total area of segments; adaxial epidermis-19-25% and abaxial epidermis-12-18% of the total area of segments (only 1% difference from the corresponding variations for individual phytomers soon after their complete expansion) (Fig. 4c,d,e).
Crude protein (CP) content and in vitro dry matter digestibility (IVDMD) varied along the leaf blades (P < 0.0001). The apical portion (segment 5) of phytomer 16, which presented the greatest lengths and smallest percentages of central rib, showed greatest IVDMD (85.8%), since it was the youngest leaf on the tiller, with   (Table 4). Smaller IVDMD values were recorded on the apical portion of small, old leaves (e.g. low hierarchical order on the tiller axis). In relation to CP percentage, greatest value was recorded on the apical segment (segment 5) of phytomer 16 (Table 5) Cluster analysis. Grouping from the Cluster analysis produced results in line with those of the PCA, which helped to their interpretation and discussion. Observing the dendrogram from left to right and using a reference cut of 0.60, it is possible to identify five well defined clusters based on the sum of the squares using the unweighted pair group method with arithmetic mean (UPGMA), according to which clusters are linked by the similarity among their elements. Clusters 1 and 3 comprised small phytomers (quadrant 3-biplot PC1 × PC2); cluster 2 comprised basal blade segments of phytomers 11 to 16 (quadrant 4-PC1 × PC2); cluster 4 comprised middle segments (quadrant 1-PC1 × PC2) and cluster 5 comprised apical segments of longer phytomers (quadrant 2-PC1 × PC2) (Fig. 7).

Discussion
Napier elephant grass is tropical forage grass with a very high dry matter production potential commonly used in intensive pastoral dairy systems 24 . Understanding the relationship between morphological, anatomical and nutritive value traits is therefore strategic for planning grazing management strategies that ensure high quality herbage. The linear relationship between total central rib mass and total blade mass ( Fig. 2) emphasizes the coordination in allocation of mass for the central rib and the leaf blade (Table 1). Longer leaves were positioned at higher levels of insertion on tiller axis, since there was no internode elongation (apical meristem elevation) and greater total blade and central rib mass were recorded. The increase in central rib mass associated with the increase in total blade length according to leaf insertion level on the tiller (Table 1), particularly for the basal portion of the blade (Fig. 1), may have been consequence of greater need to structural support to maintain the erect architecture of leaves with an appropriate angle to intercept the incident light 12 , since that the central rib is the main leaf rib and, for some tropical grasses, it usually provides additional structural support 16 . Wheller et al. 17 , studying six varieties of commonly cultivated forage sorghum, reported that the central rib represented 21 to 28% of the total leaf blade weight, smaller values relative to those recorded in this study (30 to 41%), most likely because of morphological and bulk differences between forage sorghum and Napier elephant grass. Another possibility is that physiologically, longer leaves have longer hydraulic path lengths, and because path length influences the water potentially required to continually conduct water, both stems and leaves have evolved tapering of major Table 3. Linear regression between segments of the leaf blade (basal, mid-basal, middle, mid-apical and apical portions) and the percentage of anatomical tissues of leaves of contrasting insertion levels along the tiller axis of elephant grass. α intercept, SE standard error, β slope. 13 12.  www.nature.com/scientificreports/ vein xylem from the base to the tip of the leaf (from the base to the tip of the stem as well), which has evolved to overcome this path length constraint 18 .
Regarding the anatomical composition of the leaves, the literature describes that large leaves, with higher insertion levels on the tiller, show greater percentage of structural support tissues and cells with thickened secondary walls than small leaves, with lower insertion levels on the tiller 8,25 . Wilson 25 , in a study with Panicum maximum var. trichoglume, reported a variation gradient on leaf tissue proportions along the tiller axis. Paciullo  The variation in leaf blade anatomy according to the insertion level on the tiller axis is a specific characteristic related to tiller development 25 that may be observed in most grasses. The unique feature of this study is that it focused on studying the anatomical variations along the entire length of each leaf on the tiller axis (contrasting   (Fig. 3a). Greatest percentage of sclerenchyma and vascular tissue were also recorded on the segments close to the base of the blades, ensuring stronger structural support (segments 1 and 2, Fig. 3b). Sclerenchyma and xylem cells are important for ensuring the erect architecture of leaves, particularly in warm and dry environments 13,14 . Greater percentages of parenchymatic sheath of the vascular bundles and vascular tissue were recorded on the middle portion of the blades (Fig. 3c), possibly because of the greater blade width and smaller percentage of central rib mass relative to the basal portions of the blade. Greater percentages of epidermis (adaxial and abaxial) were observed in the apical portion of the blades (segments 4 and 5, Fig. 3d,e). The epidermis is comprised of live, vacuolated cells with the function of restricting water loss and, as a result, does not have intercellular space 26,27 . In spite of the small anatomic composition variation along the length of the leaf blades, the distribution of anatomical tissues was relatively stable along the length of leaves of different sizes and levels of insertion on the tiller axis, i.e. the same pattern of variation in anatomical composition was observed for all phytomers regardless of their hierarchical order of development. These findings support the use of protocols based on measurements of the middle portion of the blades only, reducing the amount of field and lab work in studies of this nature. When leaf age was taken into account (harvest of all leaves only when leaf 16 had completed expansion), the results were similar to those of leaves harvested individually soon after their full expansion (Fig. 4). This is probably because after complete expansion the anatomical composition of leaves is already defined and does not change with time, resulting in similar relative contribution of each anatomical tissue regardless of leaf age 25,28,29 . There are evidences in the literature consistent with our findings in which the effect of age and climatic factors on leaf anatomical composition were also small 19,30 . Wilson 25 , in a study with Panicum maximum var. trichoglume, reported no variation in the proportion of anatomical tissues with increases in leaf age, but sclerenchyma cell wall was thicker, particularly in leaf sheaths positioned at the apical portion of the tillers (e.g. higher insertion levels). Similar results were reported for Brachiaria decumbens, Melinis minutiflora and Cynodon sp. (Tifton-85) 8 .
The central rib influenced IVDMD and CP along the length of the leaf blade, since when chemical analysis was performed on blade material without the central rib (e.g. taking into account only the anatomical composition of the blade), differences were too small and did not explain the differences in IVDMD and CP of whole leaves. Variability in IVDMD among C 4 species is more dependent on the rate and extension of degradation of tissues like mesophyll, parenchymatic sheath of the vascular bundles and sclerenchyma than the amount of the  Table 4). The structural support assured by the central rib to the blade is mainly provided by the vascular tissue (xylem) and sclerenchyma 5 . These tissues are formed by cells from the thickened secondary wall and are the main responsible for the reduction in forage nutritive value 15,16 , since they form a solid multicellular block in the rumen that results in large particles of difficult digestion because of lignification and associated restrictions to colonisation by rumen microbes 15,32 . As the central rib usually contains large percentage of vascular tissues and sclerenchyma, its digestibility is significantly smaller than the blade 17,33 . As a result, the basal segments of leaves of higher insertion levels on the tiller (e.g. bigger leaves), because of their larger mid rib mass, show smaller IVDMD than the middle and apical in spite of their greater percentage of mesophyll, one of the most easily digested plant tissues 34 . Although leaves 8, 9, 10 and 11 showed smaller IVDMD on the apical portions, the average IVDMD for the whole leaf blade of those phytomers was 82.22; 83.05; 82.08 and 82.12%, respectively. For the phytomers of higher level of insertion on the tiller (phytomers 12, 13, 14, 15 and 16), the average IVDMD values of middle, mid-apical and apical segments (segments 3, 4 and 5, respectively) were 82.71; 82.24; 82.63; 82.68 and 82.82%, respectively, similar to those of phytomers of lower level of insertion despite these being older. The smaller IVDMD of the apical segments of leaves 8, 9, 10 and 11 may have been consequence of their advanced maturity, since the stage of development is the most important factor influencing the nutritive value of forage grasses 35 . The apical segment of the blades for phytomer 8 and part of the blades for phytomer 9 showed evident signs of senescence like yellowing, darkening and tissue degradation due to remobilization of nutrients to other growing points, a condition that is well known for reducing the nutritive value of leaves 35,36 . On the other hand, phytomers 5, 6 and 7 were completely senesced at the time of harvest when leaf 16 completed expansion. If only the green portion of the blades is considered, with no export of nutrients to growing points (i.e. leaf senescence), it is possible to verify the negative impact of the central rib on the digestibility of leaf blades.
The greatest CP value was recorded on the apical segment (segment 5) of phytomer 16 (Table 5), and did not differ from those segments with smaller percentages of central rib mass and greater values of SLA. These findings are in line with those of Lucena et al. 23 , which indicated positive correlations between CP and SLA (r = 0.52; P = 0.0010) on leaf blades of Paspalum spp.. The difference in CP detected on segment 5 of phytomer 10 may have been consequence of leaf maturity stage, similarly to all segments of phytomers 8 and 9, since cell wall components and fibre increase and digestibility and crude protein decrease as leaf age increases 7,14 . Protein, along with energy, is the most required nutrient by ruminants, and the smaller CP content observed in this study (11%) is significantly greater than the minimum recommended of 7% for adequate ruminal fermentation 37 . Similar to the IVDMD results, central rib exerted a negative effect on CP content, possibly because of the reduction in blade proportion, where tissues with greater nitrogen content like mesophyll and parenchymatic sheath of the vascular bundles are present.
The findings of the multivariate analysis (Cluster and PCA) corroborated and integrated the results from single variable analysis (ANOVA), since they generated clusters and principal components according to the level of insertion of phytomers on the tiller axis and their structural support (e.g. blade length associated with central rib distribution along the blade), relating morphological, anatomical and nutritive value characteristics. On the PCA-derived biplots, quadrants are inversely correlated (vectors in opposite directions). Large vectors indicate large variability in the data set, and proximity among vectors indicates how strongly related they are (Figs. 5, 6). The basal portions of blades from phytomers of high level of insertion on the tiller axis were characterised by structural support variables (SCL; CRM and CRW), in line with the morphological data (greater central rib mass). Additionally, they were associated with greater mesophyll percentage, probably because they were thicker leaves (smaller SLA) and, consequently, of lower nutritive value (Fig. 5, quadrant 4). According to the literature, SLA is usually related to leaf anatomy, with large SLA associated with thinner cuticles, epidermis and mesophyll 38,39 . The middle portions of the blades corresponded to the regions where blade width was greater, consequently with greater concentration of vascular bundles resulting in greater blade mass relative to the entire length of the leaf (Fig. 5, quadrant 1). The apical portions were thinner, with larger SLA and higher nutritive value, possibly because of their smaller central rib mass (Fig. 5, quadrant 3).
In general, phytomers of lower insertion level on the tiller (smaller leaves), although older, were characterised by nutritive value variables (IVDMD, CP and ASH), TEpi and SLA. In spite of the slow and partial digestion of the epidermis from tropical grasses 15 , this variable was associated with regions of higher nutritive value, in line with the findings of Wilson et al. 31 , according to which epidermis cells were rapidly and extensively degraded for the majority of the Panicum species studied. However, significant correlations between digestibility and epidermis percentage were not found 31 . The bulk of the characteristics associated with the smaller phytomers indicate the strong negative impact of the central rib on the nutritive value, particularly over the IVDMD, since the anatomical composition of the blades did not change with leaf age 5 .

Conclusions
Leaf blades of Napier elephant grass show negative relationship between central rib and nutritive value, highlighting the importance of the central rib in determining the nutritive value of the herbage. The apical portions of the blades have similar digestibility to small leaves regardless of leaf chronological and physiological age, indicating that grazing management that favours the removal of the apical portion of leaves (moderate/lenient defoliation regimes) ensures the harvest of high nutritive value herbage by the grazing animals.

Methods
All procedures were approved by the Animal and Environment Ethics Committees of the University of São Paulo, College of Agriculture "Luiz de Queiroz" (USP/ESALQ). All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. To avoid soil water deficits, a drip irrigation system was installed in the area used for Experiment 1 and a sprinkler irrigation system was available in the area for Experiment 2. Irrigation in both areas was carried out according to records of precipitation, average air temperature and evapotranspiration. On rainy days, precipitation was recorded and taken into account in calculations for irrigation as a means of ensuring that plants were not submitted to either deficit or excessive soil moisture. One day before planting of the experimental area (290 m 2 ), planting pits were opened and the desiccated vegetation around them removed. Planting material (stem cuttings with viable lateral buds) was harvested at an 850 m 2 pasture of well-established Napier elephant grass 41 . Stems were fractioned in one-node pieces (one single axillary bud) discarding the basal and the apical portions of the stems to ensure vigorous sprouting from the planting material. Ten buds were placed in each pit at 5 cm depth, covered with soil and gently compressed by hand. The distance between pits in lines was 1.5 m and between planting lines was 2 m, in order to obtain the desired spaced-plant layout. Two weeks after planting, plants were thinned leaving one plant per pit.

Experimental site.
Treatments corresponded to phytomer order along the tiller axis and the experimental design was a randomised complete block, with four replications. Plants within blocks were randomised using the statistical package SAS (Statistical Analysis System, v. 9.0).
Sampling followed the ontogenetic programme of plants, beginning with tiller 1, which corresponded to the anatomical evaluation of the 8th leaf (phytomer 8); tiller 2, which corresponded to the anatomical evaluation of the 9th leaf (phytomer 9) and the morphological evaluation of the 8th leaf; tiller 3, which corresponded to the anatomical evaluation of the 10th leaf (phytomer 10) and the morphological evaluation of the 9th leaf in this sequence until full expansion of the 16th leaf (phytomer 16) for morphological evaluation (tiller 10), totalling 40 tillers (4 tillers for anatomical characterisation of the 8th expanded leaf + 32 tillers for anatomical (9th to 16th leaf) and morphological (8th to 15th expanded leaf) evaluations + 4 tillers for morphological characterisation of the 16th expanded leaf). At each harvest, leaves were carefully removed from the tillers, identified and preserved with ice until processing in the laboratory. Sampled tillers were removed from the experimental area. In order to evaluate the effect of leaf age on the deposition of support tissues, all leaves from 8 tillers (4 for anatomical evaluations (tiller 11) and 4 for morphological evaluations (tiller 12)) were collected at a single harvest when the 16th leaf completed expansion following the same procedure described for each leaf separately. Twenty additional plants were grown (five per block) to ensure that all leaves would be harvested as planned, but they were not necessary.
Leaf anatomy. In the laboratory, the leaf blade was cut at the ligule, its length was measured (distance between the tip of the leaf and the ligule) and fractionated in five segments of similar length designated as: (1) basalclosest portion to the insertion on the tiller; (2) mid-basal-middle portion closest to the basal; (3) middlemiddle portion of the blade; (4) mid-apical-middle portion closest to the apical; (5) apical-portion closest to the tip of the leaf. Each of the five segments were fragmented in 1-cm cuts and stored according to methodology described by Johansen 42 . Sample dehydration was carried out using a progressive alcoholic series with tertiary butyl alcohol 43 , and fragments infiltrated with paraffin and subsequently with paraplast. In sequence, fragments  ). Initially, the whole crosssection area projected on the video was measured (STotal). Next in the measurement sequence were the areas of adaxial (EPIada) and abaxial (EPIaba) epidermis, parenchymatic sheath of vascular bundles (PSV), vascular tissue (VT-including xylem, phloem, mestome sheath and pericyclic fibers 27 ) and sclerenchyma (SCL). The mesophyll area was calculated as the difference between STotal and that of all the other tissues, therefore including airspace. Measurements were made in µm 2 and the results expressed as percentage of total area.
Leaf morphology. For the morphology measurements the leaf was cut at the ligule, its length was measured (distance between the tip of the leaf and the ligule) and fractionated in ten segments of similar length. At the mid portion of each segment the fragment width (distance between the opposite borders) was measured in millimetres. In the sequence, the central rib from each segment was removed using a scalpel and its width and length were also measured. The fragment parts (central rib and blade tissue) were passed through a LAI-3100 leaf area integrating device (LI-COR) and put to dry in a forced draught oven at 55 °C until constant weight. The results were used to calculated whole segment mass (mg) and specific leaf area (SLA-cm 2 .mg −1 ). . The next steps followed the same protocol used in Experiment 1. Six buds were placed in each pit at 5 cm depth, covered with soil and gently compressed by hand. The distance between pits in lines and between lines was 1.0 m (Fig. 8).

Experiment 2 (nutritive value
A total of 1,320 plants were cultivated. These were divided in three homogeneous blocks with 440 plants each. The number of plants per block was dimensioned to provide the necessary amount of dried and ground samples for the nutritive value analysis. Twenty days after planting, plants were thinned leaving one plant per pit. Treatments corresponded to phytomer order along the tiller axis and the experimental design was a randomised complete block, with three replications. Plants within blocks were randomised using the statistical package SAS (Statistical Analysis System, v. 9.0). Weed, pest and disease control was the same as for Experiment 1.
Sampling was carried out when the 16th leaf (phytomer 16) complete its expansion and exposed the ligule. Main tillers had all their leaves identified with permanent marker before harvest. The leaves were stored in ice and taken to the laboratory. Processing involved removal of the sheaths and blades stored in freezer for future segmentation. Leaves 5, 6 and 7 were discarded because they were in advanced stage of senescence and decay, leaving leaves 8 to 16 for the analysis.
Nutritive value. After all field work was finished, leaf blades had their length measured (distance between the tip of the leaf and the ligule) and were fractionated in five segments of similar length, as described for leaf anatomy measurements in Experiment 1 (i.e. chemical analyses included all tissues of the leaf blade). These were pooled into a composite sample per leaf hierarchical order, totalling 135 samples (9 leaves × 5 segments × 3 blocks). These were put to dry into forced draught oven at 55 °C until constant weight.
After drying, because the apical portion of the leaves was too delicate, the dried material was ground in a micro mill with a 1 mm sieve (Wiley Mill, Thomas Scientific, Philadelphia, PA, USA) as a means of reducing dry matter losses. Ground samples were subjected to the following chemical analysis: in vtiro dry matter digestibility (IVDMD), using the artificial rumen fermentation device DAISYII from ANKOM Technology Corporation 45 ; total nitrogen (N total ), determined by the Dumas combustion method using the Leco FP 528 System (Leco Instruments Inc., St. Joseph, MI, USA); and ashes (ASH) determined according to Silva & Queiroz 46 . Crude protein (CP) was calculated as N total × 6.25. For the in vitro trial, rumen liquid was collected from one rumen-cannulated Nellore steer fed only with Tifton-85 (Cynodon dactylon spp.) haylage.

Statistical analysis.
Leaf blade morphology and anatomy data were initially analysed using descriptive statistical analysis (means and standard error of the mean). Leaf blade total mass and central rib total mass were obtained by adding values for all ten segments from each leaf and were subjected to a regression analysis as a means of identifying the correlation between these two variables. Regression was performed using the procedure PROC REG of SAS (Statistical Analysis System, v. 9.0).
Nutritive value data were subjected to ANOVA using the procedure PROC GLM of SAS (Statistical Analysis System, v. 9.0), and means compared by Tukey test (P < 0.05). The statistical model considered phytomer and block as fixed effects: www.nature.com/scientificreports/ where Y ij : average response of the jth phytomer in the ith block; µ : mean; B i : block i; F j : phytomer j; ε ij : random error ~ NID (0;σ 2 ε ). All assumptions for these analyses were verified as homogeneity of variances, error normality, outliers, and need for transformation by Box-Cox 47 .
A Principal Components Analysis (PCA) was carried out with the objective of integrating the leaf morphology, leaf anatomy and nutritive value results. The data set was comprised of the following response-variables: Morphology-blade length, segment length, blade width, central rib width, total width (blade + central rib), blade mass, central rib mass, total mass (blade + central rib) and specific leaf area; Anatomy (percentage of the cross sectional areas for each tissue)-mesophyll, vascular bundles (parenchymatic sheath of vascular bundles + vascular tissue), total epidermis and sclerenchyma; and Nutritive Value-crude protein, in vitro dry matter digestibility and ash (data from Experiment 2 when all leaves were harvested at full expansion of leaf 16). Because leaf blades were fractioned in ten segments for morphology measurements and five segments for anatomy and nutritive value measurements, segments from the morphology measurements were combined two-by-two to balance segment number from all measurements (1 + 2, 3 + 4, 5 + 6, 7 + 8 and 9 + 10 to form 1, 2, 3, 4 and 5, respectively). As a result, all response variables had blade segments associated with them. PCA was performed using the procedure PROC PRINCOMP and the biplots were generated using the procedure PROC PRINQUAL of SAS (Statistical Analysis System, v. 9.0).
Cluster analysis was performed using the same PCA data set with the unweighted pair-group method with arithmetic mean (UPGMA) using the procedure PROC CLUSTER of SAS (Statistical Analysis System, v. 9.0) as a means of identifying phytomers groups according to the similarity of their elements.

Data availability
The datasets generated during and/or analysed during the current study may be available from the corresponding author upon request.