Differences also exist between dicot and grass cell walls in the thicker, secondary cell walls that accumulate upon growth cessation. Secondary walls are especially abundant in three polymers—cellulose, xylan, and the phenylpropanoid-derived polymer, lignin. As explained above, the xylan of grasses is modified by arabinose, which is nearly lacking in dicot xylan (Scheller et al. 2010). Instead, dicot xylan is modified by glucuronic acid, which is less frequent in grass xylan (Fig. 2). As in primary walls, arabinose residues of xylan in grass secondary walls are acylated with ferulic acid and p-coumaric acid. Indeed, the presence of ether bonds between feruloyl esters from glucuronoarabinoxylan and lignin suggests that feruloylated arabinose could act as a nucleation site for lignin formation in grasses (Bunzel et al. 2004). Dicot xylan has a 5-sugar oligosaccharide at the reducing end; however, this sequence has not been identified in grasses, despite concerted efforts (Scheller et al. 2010) (Fig. 2).

Due to its abundance and high energy content, lignin represents an attractive byproduct of biochemical biofuel production for conversion to electricity, thermochemical biofuels, or higher value chemicals. Lignin is deposited in interstices of the cell wall during secondary development (Terashima et al. 2004). Xylem and schlerenchyma cells are particularly rich in lignin, but in switchgrass internodes closer to the plant base, all cell walls stain for lignin (Shen et al. 2009). The working model for lignin polymerization is that monomers are released into the wall where they are enzymatically oxidized to form radical ions that form covalent bonds with each other and other nearby alcohols (Boerjan et al. 2003). Grass lignin varies from the lignin of dicot walls in that it includes significant amounts of p-hydroxyphenyl (H) subunits in addition to the typical guaiacyl (G) subunits and syringyl (S) subunits also present in dicot walls (Fig. 3) (Vogel 2008). Each of the lignin subunits has a different number of potential bonding sites, with four, three, or one bond observed, respectively, in model studies (Fig. 3) (Boerjan et al. 2003). Thus, higher amounts of H and G lignin have the potential to lead to a more branched lignin structure. Nonetheless, the amount of H lignin in switchgrass stems, for example, has been found to be low, with typical ratios of 0.1:1.0:0.8 (H:G:S) in mature stems (Shen et al.

2009) . Lignin is the major polymer that blocks digestion of cell walls, with the lignin of grasses being no exception. Switchgrass stem lignin content is inversely correlated with digestibility (Shen et al. 2009).

A further distinguishing characteristic of grass cell walls is that grass monolignols are acylated by p-coumaric acid at the alcohol at the end of the propanoid "carbon tail" (Fig. 3) (Ralph 2010). Bioechemical analysis

suggests that the function of these modifications may be to indirectly enhance lignin polymerization (Ralph 2010). Though p-coumaric acid is readily oxidized to its radical, bonds with p-coumaroyl pendant groups have not been observed in planta (Ralph et al. 1994). Rather, p-coumaroyl esters may act as "radical catalysts", rapidly passing the radical to sinapyl alcohols, potentially facilitating lignin polymerization (Takahama et al. 1994; Ralph 2010). In switchgrass stems, approximately 20% of monomers are esterified to a hydroxycinnamate residue (Yan et al. 2010).

Generally speaking, young tissues that are richer in primary walls have long been recognized as being easier to digest than more mature tissues. This trend has also been observed for switchgrass stem tissues (Shen et al. 2009). Indeed, overexpression of a microRNA that represses the juvenile to mature transition, miRNA156 or Cg1, fixes the resulting switchgrass in a juvenile state and increases saccharification yields (Chuck et al. 2011). Somewhat unexpectedly, the most dramatic change in weakly overexpressing Cg1 switchgrass is a 250% increase in starch, which is made of valuable glucose residues, resulting in an overall increase in sugar yield in the presence of starch degrading enzymes of ~200% (Chuck et al. 2011). Thus, the improved sugar yields with young tissues may not be only a consequence of changes in cell wall quality, but of other physiological properties. The absence of simple correlation between lignin content and recalcitrance is also demonstrated by the fact that leaf material is more recalcitrant than stems, despite lower amounts of lignin in leaves versus stems (Fu et al. 2011).

Biomass Content Variation with Environment and Genotype

In addition to varying across development and among organs, switchgrass biomass content varies with environmental conditions and natural, selected, and engineered genetic variation. Genetic engineering to improve cell wall quality is extensively discussed below in the context of what is known about grass cell wall synthesis and regulation. This section briefly analyzes observations of other factors that influence cell wall quality. With the implementation of modern genetic methods, the authors predict that these observations may soon be moved toward understanding and harnessing molecular mechanisms.

Studies aimed at describing environmental effects on switchgrass biomass quality are few. Schmer and colleagues recently reported the year-to-year and site-to-site variability observed for their well-studied, large-scale switchgrass experiment in the Northern Great Plains (Schmer et al. 2012). In that experiment with ten upland switchgrass fields scattered in eastern North Dakota, South Dakota and Nebraska, significant variation existed among plots and among harvest years for theoretical ethanol yield in liters per megagram (L per Mg) (Schmer et al. 2012), as determined by near infrared spectroscopic analysis (Vogel et al. 2010). Note that this calculation of theoretical yield is based only on cell wall content, not structure. The year-to-year differences were as much as 20%, with theoretical ethanol yields being especially low in the establishment year in some fields, though more typically they varied by less than 10%. Though genetic differences among stands cannot be ruled out, theoretical yields per unit mass were lower in more northerly fields compared to more southerly fields, which is another possible indication of environmental influences, such as temperature, on cell wall content at maturity. Similarly, lower sugar extractability for switchgrass grown at higher latitudes was also observed in another less extensive study, though in that case the ecotypes were also different (Kim et al. 2011). Schmer and associates (2012) observed that drought increased biomass xylose content, which typically decreased ethanol yields. On the other hand, within-field variation among subsamples taken in a single year was typically only 2 to 3%, suggesting that biomass content is not particularly sensitive to the soil and moisture variability that might occur within a single field (Schmer et al. 2012). From a biofuel production perspective, this suggests that within field bail-to-bail variation in switchgrass biomass content is not expected to be large.

Compared to characterizing environmental variability, relatively more work has been done to measure the influence of genetic variation, i. e., cultivar-to-cultivar or genotype-to-genotype variation, on switchgrass cell wall quality. Molecular analysis has found that most switchgrass cultivars are relatively poorly defined genetically, with more variability observed within a particular cultivar than among cultivars (Cortese et al.

2010) . Similarly, several studies have also observed little difference among cultivars in terms of biomass quality when grown side-by-side. Out of eight cultivars grown in a randomized trial in Alabama, differences in biomass quality were observed for lowland vs. upland ecotypes, but not for most parameters within each ecotype (Sladden et al. 1991). In that study, lowland types showed both higher lignin and cellulose compared with the uplands. Similarly, in a study of small plots of 20 switchgrass genotypes grown in Iowa, researchers observed few statistically significant differences in biomass content among cultivars over three growing season (Lemus et al. 2002). The lowland cultivars, Alamo and Kanlow, had ~12% less lignin and ~14% less ash compared with the other, mostly upland cultivars in the trial (Lemus et al. 2002). Cellulose did not vary significantly among cultivars, but, as with other studies, significant biomass quality variation was observed in all cultivars depending on the year and harvest date. Another small — scale trial of lowland cultivars in Georgia revealed no major differences in internode cell wall content or structure (Yan et al. 2010). Finally, in the large-scale experiment by Schmer et al. (2012), though biomass glucose and xylose content differed significantly among the three upland cultivars grown in three sites in Nebraska, differences in theoretical ethanol yield per Mg among the cultivars only appeared at one site. An absence of differences among cultivars is not surprising given the genetic variability of switchgrass that might lead to apparent homogenization in populations of traits with potential fitness effects, such as those related to cell wall function.

However because cell wall content is genetically determined, recurrent selection for switchgrass genotypes with divergent digestibility has been successful. In particular, Vogel, Sarath and colleagues at the University of Nebraska have developed and characterized upland switchgrass populations with decreased and increased digestibility relative to the parental population (Hopkins et al. 1993). The realized heritability coefficient for digestibility was 0.31 (Hopkins et al. 1993). In general, plants bread for high digestibility exhibit lower lignin content and vice versa (Sarath et al. 2008). Low-lignin plants appear to possess less lignified stem cortex cells and give ~40% higher actual ethanol yields per gram compared with divergently selected, high — lignin genotypes (Sarath et al. 2011). These genotypes show that genetic improvement for biomass quality traits is possible through breeding. Such studies would be accelerated with the use of molecular markers (reviewed in Bartley et al. 2013b). Mapping the loci responsible would provide specific insight into control of cell wall synthesis in switchgrass. Unfortunately, in the case cited here, this is non-trivial because the divergent genotypes are octoploid, though a similar experiment has been conducted for a tetraploid population (G. Sarath, personal communication). Nonetheless, these results provide insight into achieving robust switchgrass plants with improved

digestibility. For example, reducing the lignin outside of the vasculature appears to be a good way to achieve improved cell wall quality.