Rhizomes from 11 diverse Miscanthus genotypes (Table 1) were dug and split from 10-year-old field grown plants during early spring in March 2022. The rhizome was taken from a field trial based in Aberystwyth, UK, that was originally planted as part of a wider European plant trial (Kalinina et al., 2017). After splitting the rhizomes were then pared back to one bud and planted into 1 litre pots using a mix of loam-based compost (John Innes No. 3) and perlite (30%) and were grown in a polytunnel until the following spring to develop a root system. Then five replicates from each genotype were carefully washed to remove perlite and compost and potted on into 10 litre pots using sieved (1 cm mesh) compost (John Innes No. 3). Plants were randomised within the polytunnel and grown until December with manual watering according to need. Air temperature, relative humidity, and Photosynthetically Active Radiation within the polytunnel were recorded over the study period of April to December 2023 (Skye Instruments, UK) (Supplementary Table 1).

From October until the biomass harvest which took place in December/January senesced leaves were collected from the plants before abscission. For the biomass harvest the pots were left to dry out and any remaining leaves on the plant were collected and separated into green leaves and senesced leaves. The stems were cut off at ~2.5 cm above the pot surface and removed. Below-ground biomass was gently separated from the dry compost by hand and the root and rhizome portions separated. The majority of the compost attached to the biomass was removed by hand by gently shaking and rubbing before the biomass was finally rinsed with water. The compost from each pot was then sieved (1 cm mesh) to capture any sizeable biomass remaining. Senesced leaf and separated below-ground biomass were oven dried to constant weight (40°C) to obtain the dry mass (g_dm) and then, in preparation for laboratory analysis, representative subsamples were milled to 2 mm (Pulverisette 15 mill, Fritsch GmbH, Germany).

Acid Detergent Fibre (ADF) and Neutral Detergent Fibre (NDF) were determined using the Filter Bag Technique and the ANKOM A220 Fiber Analyser (ANKOM Technology, USA) as described in ANKOM (2017a) and ANKOM (2017b). Results were expressed exclusive of residual ash. The hemicellulose content was determined by subtracting the ADF value from the NDF value. For Acid Detergent Lignin (ADL), the ADF residue was treated with 72% sulphuric acid for 3 h in a DaisyII Incubator (ANKOM Technology, USA) leaving behind the lignin fraction measured on an ash-free basis as described in ANKOM (2022). The cellulose content was determined by subtracting the ADL value from the ADF value. Analysis of the percentage carbon (C) and nitrogen (N) content was carried out on ball-milled (Labman automated preparation system) senesced leaf, rhizome and root samples using an ANCA-SL elemental analyser (Sercon Ltd, UK).

All means estimated with ± values reflect the standard error of the mean (SEM). Data analysis was carried out in R version 4.2.3 (R Core Team, 2023). Genotypic differences in plant traits (of biomass dry matter; C, N, lignin, cellulose, and hemicellulose contents; and ratios of C:N, lignin:N, and cellulose: hemicellulose) were tested using separate one-way ANOVAs for each biomass type (senesced leaf, rhizome, and root) and trait. The following datasets were log transformed to improve the normality of model residuals: root biomass dry matter; leaf, rhizome and root N content; rhizome cellulose content; leaf C:N ratio; and leaf cellulose:hemicellulose ratio. A square root transformation was performed on rhizome cellulose content data, and an inverse transformation on rhizome cellulose:hemicellulose ratio data. Tukey HSD post hoc tests (package ‘multcomp’, Hothorn et al., 2008) were used when a significant ANOVA result (p<0.05) was obtained.

To explore trait relationships with soil organic carbon (SOC, Mg ha^-1) and Miscanthus derived soil carbon (C₄-C, Mg ha^-1) a dataset of soil C stocks from under the same, but 10-year-old field grown plants (Holder et al., 2025), used to source the rhizome for this study, was used with the plant trait results obtained in the polytunnel experiment. Full methodology for soil sampling is available in Holder et al., 2025. Akaike’s information criterion (AIC) was used for the selection of best fit linear models with fixed factors of soil sample depth increment (0-10, 10–20 and 20–30 cm), lignin content, cellulose content, hemicellulose content, C:N ratio, lignin:N ratio, and cellulose:hemicellulose ratio (R packages “nlme” (Pinheiro et al., 2017) and “MuMIn” (Barton, 2023). Separate models were used for each biomass and soil C type (SOC and C₄-C). A square root transformation was used on C₄-C data to improve model residuals. A Pearson’s correlation statistic was obtained for soil C and the fixed factors identified as improving model fit.

When combined with soil C data from the same genotypes grown in the field, lignin content and the lignin:N ratio were shown to be useful predictive factors (along with soil depth) in best fit model selection suggesting the importance of these traits for soil C. The other traits considered (C:N ratio, cellulose, hemicellulose, and cellulose:hemicellulose ratio) did not show any relationship with the soil C data.

Senesced leaf lignin:N ratio, along with soil depth were selected as important factors for predicting soil C₄-C, explaining 86% of the model variance (Figure 2). However, no significant correlation was found between the lignin:N ratio and C₄-C (r₃₁ = -0.22, p = 0.22; r₃₁ = -0.29, p = 0.11; r₁₉ = -0.20, p = 0.39, for soil depths 0–10 cm, 10–20 cm and 20–30 cm, respectively). None of the leaf traits included were found to be important in predicting SOC.

Rhizome lignin content and soil depth were key factors in relation to SOC, although only predicting 55% of the model variance. However, in addition to this, rhizome lignin content was positively and significantly correlated to SOC at the 10–20 cm soil depth (Figure 3) (r₃₁ = 0.02, p = 0.91; r₃₁ = 0.34, p = 0.05; r₁₉ = 0.14, p = 0.54, for soil depths 0–10 cm, 10–20 cm and 20–30 cm, respectively). No association was found between the rhizome traits considered and soil C₄-C.

Root lignin content and the lignin:N ratio, along with soil depth, were also identified as important factors in predicting soil C₄-C, with this combination of factors explaining 86% of the model variance. Root lignin content was negatively and significantly correlated with soil C₄-C for the 0–10 cm soil depth (Figure 4A), (r₃₁ = -0.42, p = 0.01; r₃₁ = 0.18, p = 0.31; r₁₉ = -0.10, p = 0.66, for soil depths 0–10 cm, 10–20 cm and 20–30 cm, respectively), but no significant correlations were found for soil C₄-C and the lignin:N ratio. However, a relationship was found between root lignin:N ratio and SOC, which along with soil depth was selected as a key model predictor explaining 56% of the model variance. The root lignin:N ratio was also negatively and significantly correlated with SOC at the 10–20 cm soil depth (Figure 4B) (r₃₁ = 0.01, p = 0.96; r₃₁ = -0.35, p = 0.04; r₁₉ = -0.13, p = 0.58, for soil depths 0–10 cm, 10–20 cm and 20–30 cm, respectively).

The results of this study provide much needed data of Miscanthus cell wall composition of senesced leaves, rhizome, and roots. Currently very little is known about below-ground biomass traits for Miscanthus genotypes, how they differ, their relevance for Miscanthus breeding, or how they could potentially impact soil C cycling.

All the Miscanthus genotypes we examined had below-ground biomass considered to be recalcitrant compared to annual biomass crops, but less so compared to woody crops, and more similar compared to other perennial herbaceous crops. For example, the root lignin content we recorded for the Miscanthus genotypes (11%-14%) is generally lower for maize (Zea Mays, 4%-10%) (Machinet et al., 2011), higher for poplar (Populus spp. ~18%) and similar for Switchgrass (Panicum virgatum, 13%) (Ferrarini et al., 2022).

For all the Miscanthus genotypes included in this study above- (senesced leaves) and below-ground (rhizome and root) biomass differed in terms of N, lignin, cellulose and hemicellulose content. The lower N content of the senesced leaves reflected the nutrient translocation from leaves into plant storage organs (Magenau et al., 2022). Significant Miscanthus genotypic variation in cell wall composition was also found in each of the tissue types.

Significant genotypic variation was found for all the senesced leaf traits explored providing scope for breeding. Differences in Miscanthus leaf trait (including nutrient content) relationships between triploid and tetraploid genotypes have been identified (Li et al., 2022) showing the potential for polyploidization for desired leaf traits. In other species cell wall traits are amenable to breeding and many studies have included GWAS to understand the genetic architectures of lignin, cellulose and hemicellulose for breeding programmes (Esposito et al., 2022; Panahabadi et al., 2022; Yang et al., 2019).

The rhizome mass recorded for M×g and Sac×Rob were similar to the higher rhizome mass recorded for the three Sac species. In the Sac species root lignin content was also generally higher than rhizome lignin whereas for the Sin species and hybrids it was similar. But with the exception of this none of the hybrids showed any other traits that could be particularly linked to the Sac or the Sin species.

In the case of studies of soil organic carbon accumulation with Miscanthus crops they traditionally occur over many years with samples taken from mature field crops between 3 to 20+ years old (Qin et al., 2016; Zang et al., 2018). Whereas, mostly, morphology and composition traits have been assessed across shorter time scales and often in pots. It is therefore worth considering if there is a reasonable expectation that the traits measured in the pot experiment will be consistent, at least in terms of rank order, with the same traits expressed over many years in the field. For some cell wall traits such as lignin many experiments have manipulated composition and genotypes engineered for high lignin phenotypes in the pot, for example, are also high lignin genotypes in the field (De Meester et al., 2022). The environment and plant age can impact lignin content (da Costa et al., 2014), for example high temperatures (Crivellaro and Büntgen, 2020) and other abiotic and biotic stresses (Cesarino, 2019) affect lignin accumulation. Small variations have been seen in contents for above-ground biomass for these genotypes when grown in different European locations (Kiesel et al., 2017) but generally the control of lignin content is moderate to highly heritable (Harman-Ware et al., 2021; Mandrou et al., 2012; Panahabadi et al., 2022). This is consistent with cellulose, hemicellulose and lignin contents analysed from a diversity panel of Miscanthus, where all composition measurements had a similar and high broad sense heritability with an average across all cell wall traits of 0.67 (Slavov et al., 2013).

Contrary to results obtained for the above-ground harvested biomass (mainly stems) (Allison et al., 2011; Hodgson et al., 2010; Kiesel et al., 2017), we did not find the senesced leaf for the common commercial clone M×g (OPM9) to be higher in cellulose or lignin compared to the Sac and Sin species. However, M×g was among the lowest in terms of senesced leaf hemicellulose content as was the case for the harvested material from other studies (Allison et al., 2011; Kiesel et al., 2017). The lignin content of senesced leaf biomass (grown in Germany and separated from field harvested material in spring) for the Rob×Sin (OPM6) and Sin (OPM11) genotypes was recorded as higher than that for the Sac species (OPM3) and M×g (Schäfer et al., 2019). similar to this, in this study we also found the Sin senesced leaf to be one of the highest in lignin content, but this was not significantly higher than either the Sac (OPM3) or M×g. Both Hodgson et al. (2010) and Schäfer et al. (2019) found harvested biomass (stems and separated leaves) from the Sin species to have higher portions of hemicellulose to cellulose compared to the Sac species and M×g. In this study we also found the content comparable to two out of the three Sacs (OPM1, OPM2) and Sac×Rob (OPM4). However, in our results M×g was similar to Sin in its proportion of hemicellulose to cellulose.

Some trait differences between above- and below-ground biomass are likely to have contrasting effects for SOC accumulation. In particular, the Sac species OPM1 had traits that could decrease decomposition i.e. high root lignin content and lignin:N ratio (compared to the other genotypes) (Abiven et al., 2005), but conversely senesced leaf for this same genotype had a 1:1 cellulose to hemicellulose ratio, the lowest lignin content, and the lowest lignin:N ratio. OPM1 leaf was also among the lowest in terms of C:N ratio, the latter being all traits that could increase decomposition (Poirier et al., 2018).

Counteracting traits were also noted within the below-ground biomass samples. The Sac species OPM3 rhizome and roots, for instance, had one of the highest C:N ratios compared to the other genotypes, a trait generally linked to slower decomposition (Nicolardot et al., 2001). But OPM3 was also among the lowest in terms of cellulose:hemicellulose ratio, a trait expected to increase decomposition (Kögel-Knabner, 2002). Diverging traits were also found between rhizome and root samples. For example, OPM1 had the lowest rhizome lignin content, but was also one of the highest in terms of root lignin content.

It could be expected that the Sin species (OPM11), with the least biomass amount, one of the highest biomass lignin contents, and a below-ground biomass cellulose to hemicellulose ratio of 1, would result in lower SOC stocks compared to the other genotypes. However, results from SOC sampling of the same 10-year-old field grown genotypes used to provide the rhizome material for this study did not show this to be the case. Following soil C sampling pre-planting and after 10 years of Miscanthus growth, long-term SOC stocks under the Sin plots were found to be similar to the other genotypes (Holder et al., 2025).

Of the traits considered here (lignin, cellulose, hemicellulose content, and C:N ratio, lignin:N ratio, and cellulose:hemicellulose ratio) only lignin and the lignin:N ratio were identified as important in predicting SOC and C₄-C. Previously high lignin concentrations in plant litter have been considered to increase SOC stocks due to the slower decomposition of complex lignin compounds by only a few organisms (Bollag et al., 1997; Hall et al., 2020). And in this study, we found higher rhizome lignin content to be associated with higher SOC stocks (for the 10–20 cm soil depth). However, recent work on lignin decomposition has shown that high lignin concentrations in plant litter can have contrasting affects for SOC stocks and does not always lead to increased SOC (Hall et al., 2020; Poirier et al., 2018). In support of this and in contrast to the relationship between rhizome lignin concentration and SOC, we identified low root lignin concentration (for the 0–10 cm soil depth) and low root lignin:N ratios (for the 10–20 cm soil depth) to be associated with higher SOC and C₄-C stocks. Low senesced leaf lignin concentration was also linked with higher C₄-C stocks. This type of result has also been found for other C₄ grass species, where in a comparison of tropical perennial grasses root lignin concentration was found to be a driver of SOC, with lower root lignin content varieties accumulating the greatest SOC (Sumiyoshi et al., 2017). The lower lignin litter decomposed faster, but the residue and associated microbial by-products increased SOC (Sumiyoshi et al., 2017).

Previous studies have found M×g rhizome to have a faster decomposition rate compared to roots which has been attributed to the rhizomes higher sugar content (cellulose and hemicellulose) and lower lignin content (Amougou et al., 2011; Beuch et al., 2000; Ferrarini et al., 2022; Ridgeway et al., 2022). This is likely to contribute to the different relationships of rhizome and root with SOC stocks that we found. The genotypes we explored all had higher cellulose contents in the rhizome compared to their roots with the exception of M×g (where it was slightly less) and the Sin species (where it was the same). Soil acidity, soil N availability, and climate also have a variable influence on bacteria and fungi involved in lignin decomposition (Thevenot et al., 2010). The low pH (5.3) of the soil in this study was optimal for fungal activity, but low for bacterial degradation of lignin (Thevenot et al., 2010).

The C:N ratio of Miscanthus root litter has also been shown to influence SOC cycling where higher substrate C:N ratios reduce microbial carbon use efficiency (CUE) and the low N limits microbial growth, requiring microbes to mine for N, thereby increasing their respiration (Ferrarini et al., 2022; Poeplau et al., 2023; Ridgeway et al., 2022). In an incubation study using Miscanthus root litter it was found that litter with a C:N ratio of 85 negatively affected change in SOC more than litter with a C:N ratio of 50 (Poeplau et al., 2023). The root C:N ratio of the genotypes in this study ranged from 44 to 79, a difference large enough to potentially influence C cycling. It should also be noted that C:N ratios of root, and rhizome in particular, vary seasonally due to nutrient translocation (Heaton et al., 2009; Poeplau et al., 2019).

Although lignin content is identified as an important component of SOM formation (Stewart et al., 2015) other studies emphasise the importance of monomeric composition of lignin, in particular the ratio of Syringyl to Guaiacyl subunits in lignin degradation (Aswin et al., 2025). The impacts of lignin recalcitrance and association with soil mineral particles and the susceptibility of lignin to degradation by microbes leading to the production of microbial necromass are both routes to the stabilisation of plant-derived C in soil. The relative importance of plant biomass and microbial necromass to the production of SOM remains a subject of debate (Whalen et al., 2022) and the relative importance and interactions of the substituent components of this complex nexus of soil, plant and microbiome (Basile-Doelsch et al., 2020) remain to be elucidated.

Plant litter cell wall composition, although a principal component of soil C cycling, is subject to a number of interacting factors that can modify its influence on SOC stocks. Therefore, the relationships with SOC and the genotypic trait differences observed here may have a more pronounced or different impact on SOC accumulation in different locations with varying soil conditions (e.g. pH, soil C:N ratio), climate (e.g. temperature, precipitation), and management (e.g. fertilisation). Whilst this paper provides much needed data on the composition and genotypic variation of Miscanthus tissue entering soil C cycling systems, the quantity of C input from the turnover of rhizome and root in the field is largely unknown for both M×g and novel hybrids and remains an important metric needed to consider the contribution of Miscanthus genotypes to SOC accumulation. Research directed at a better understanding of the interaction of the traits under different field conditions, and the relevant impacts on various longer and shorter-term soil C pools would also aid in identifying trait preferences for the cell wall ideotype targeted to increased soil carbon.

Our research demonstrates useful variation in the composition of above- and below-ground tissues within a small population of Miscanthus genotypes. Some of the variation is likely to be counteracting for example the genotype OPM3 had within the same tissues a C:N ratio likely to reduce decomposition and a cellulose:hemicellulose ratio expected to increase decomposition. This may explain in part the general similarity between SOC stocks in the field trial and suggests it may be more informative to identify genotypes that accumulate SOC quicker i.e. over a shorter time span. The sometimes inconsistent effects of plant inputs when many other soil related factors are similar may be explained by complex interactions such as the level of soil carbon saturation, the unsatisfied ability of different soils to store carbon (Castellano et al., 2015). The field used in the SOC study is representative of the largest category of agricultural land classification in the UK (Class 3) (Welsh Govt., 2019), and therefore is quite broadly applicable, but it would be useful to examine if variation in SOC within different soil types and or classifications give similar results. Previously we showed that high yielding Miscanthus genotypes did not incur a soil carbon penalty (Holder et al., 2025), which is good news for breeding programmes that focus on above-ground harvestable yield. If in addition to harvestable yield breeding programmes wish to maximise carbon sequestered into soil our results suggest a high SOC accumulating ideotype includes low root lignin and lignin:N ratio and high below-ground biomass, all of which represent significant components in our models relating traits to SOC.