Limitations, considerations, and the impact of cell wall structure on S/G and associated methodologies
A comparison of S/G ratios determined by four different methods for the poplar biomass analyzed is provided in Fig. 3 and Supplementary Table 1. Generally, thioacidolysis analyses resulted in higher S/G ratios and the Manders ssNMR method plateaued at 1.1, whereas HSQC and py-MBMS data yielded the most similar S/G ratios for the poplar and corn stover samples. The deconvoluted ssNMR spectra provided S/G measurement improvements over the Manders method to more carefully reflect S/G values determined by HSQC and py-MBMS.
Lignin is not evenly distributed within a plant. Past studies have shown that almost 75% of the lignin in hardwood is contained within fiber cells, containing predominately S type lignin, and only 25% of lignin is contained in vessel and ray cells [70]. Lignin concentration is also highest in the middle lamella of the fiber cells, but the secondary cell wall contains a higher concentration of S monomers [21, 71]. High concentrations of G units are present in the middle lamella, which may be more condensed, containing a higher proportion of the total C-C linkages present that would not be released during thioacidolysis. Since S type lignin tends to be dominant in the thicker, less dense secondary cell wall and would contain more β-O-4 linkages than G units, S units could be released more effectively. Overall, monomers, and therefore condensed structures, are not evenly distributed among plant tissue types or across cell wall layers. One can reasonably assume that G monomers may not be effectively released even at low S/G ratios. This could explain why the S/G ratio measured by thioacidolysis trends high, agreeing with previous observations regarding chemical degradation methods [23, 25, 27].
Similarly, since py-MBMS is only capable of releasing and detecting lignin monomers and dimers after breaking thermally labile linkages, it may also be biased towards the analysis of moieties bound by β-O-4 linkages, particularly S-lignin monomers. If S/G ratios measured by py-MBMS were heavily biased, impacted, or correlated with β-O-4 linkages, in addition to being positively correlated with S content, then the total lignin content estimated would also be influenced such that the higher lignin content would trend with S/G. Here, the lignin content of the poplar or entire biomass sets determined by py-MBMS did not correlate strongly with S/G and lignin content estimates did not correlate strongly with thioacidolysis yields. Therefore, and in addition to strong correlation with HSQC S/G which may be less biased although is otherwise only semi-quantitative, py-MBMS S/G determined by the traditional method could be an accurate representation of S/G ratios in biomass, pending other linkage or monomeric anomalies not explored here that may otherwise impact the data. However, given the relatively small sample size of this data set and since the lignin values determined by py-MBMS did not all consistently align with solid state aromatic NMR spectral features, it may be necessary to interpret some of these correlations with caution.
Several factors contribute to the error associated with the S/G ratios measured by 1H-13C HSQC volume integrations. HSQC experiments are generally not quantitative due to the way the experiment is typically performed. Cross-peaks in a 1H-13C HSQC experiment arise from polarization transfer through one-bond J-coupling to correlate protons directly bonded to a carbon. While this provides tremendous information qualitatively, the intensity of the cross peaks is impacted by both heteronuclear and homonuclear coupling constants as well as T1 and T2 relaxation. A single 1H-13C HSQC experiment is generally optimized for one 1H-13C coupling constant; usually an average value of 145 Hz is chosen to capture both aromatic (~ 160 Hz) and aliphatic (~ 120 Hz) environments in an effort to optimize the number and intensity of the observed cross peaks. Additionally, relaxation is not accounted for in the same way as 1D experiments, where a delay of 5 x T1 is employed. These delays make an HSQC experiment untenably long. The compounding effects of differing coupling constants and relaxation rates makes the cross-peak intensities for the poplar studied here semi-quantitative, meaning only relative amounts of lignin units or linkages can be compared between samples.
13 C ssNMR analysis of lignocellulosic biomass cell walls
Solid-state NMR data were processed in two different ways (Manders et al. and spectral deconvolution or peak fitting) to extract S/G estimates directly from the in-tact biomass. Results from both data processing methods trend correctly, but it appears the Manders subtraction method undercounts S/G ratio compared to spectral deconvolution. Like thioacidolysis, py-MBMS, and gel-state HSQC analyses, there are issues related to analyses of S/G by solid-state 13C NMR methods. First, broad overlapping lines and overall poor resolution of 1D 13C solid-state NMR data poses inherent challenges, especially if detailed analyses of lignin composition are desired. Second, due to the low sensitivity of NMR in general, compounded with the low (1.1%) natural abundance of 13C, a single CP-MAS experiment usually requires long acquisition times. Sample throughput cannot possibly compete with some other analytical techniques used to obtain S/G estimates. Additionally, it is widely known that routine cross polarization NMR data is not inherently quantitative. In addition to experimental choices like magnetic field strength and rotor spinning speed, differences in cross polarization rates (TCH), spin lattice relaxation times (T1), and spin lattice relaxation times in the rotating frame (T1ρ) for carbons in different chemical environments can affect their relative intensities [72, 73]. However, in our experiments, differences in CP rates can be neglected because 1) we operated at a reasonably low (50 MHz) 13C Lamour frequency and low (6900 Hz) spinning speed such that the CP condition was quite robust, and 2) the S- and G-lignin carbons of interest used for analysis (respectively at 153 and 148 ppm) are both are non-protonated quaternary aromatic carbons in similar dynamic environments. In support, identical correction factors to adjust for variations in cross-polarization kinetics were found by Davis et al. for these two signals, confirming this assessment [64]. Therefore, careful experimental set up and data analysis can ensure that reproducible S/G ratios are determined counting all S and G units using the solid-state 13C NMR method. That said, researchers operating at higher magnetic fields and faster spinning speeds may need to take more precautions.
It is clear from comparing S/G ratios determined using the Manders method with spectral deconvolution that the procedure in which the data is processed can significantly impact results. To explain the difference, we hypothesize that the method proposed by Manders of subtracting out the softwood-derived G-lignin profile from S- and G-rich hardwood spectrum is somewhat flawed. As can be seen in Fig. 1b, in softwoods there exists a minor downfield shoulder in the G-lignin spectrum near 153 ppm. Based on multiple reports in the literature[67, 74–77] this signal can be assigned to G-lignin ring carbons at the 3, 4 and even 5 position depending on if the guaiacyl unit is or is not etherified, type of inter-unit linkage present and if carbon-oxygen condensation has occurred at the C5 position (labeled G’3,4,5 in Fig. 1b). The abundance of the G-lignin downfield shoulder near 153 ppm seen in pure-G softwoods is therefore unlikely to match the same G-lignin profile in S- and G-rich hardwoods. As a result, subtracting a G-lignin profile from hardwood 13C data could unintentionally subtract signal that is truly from S-lignin. This likely explains why S/G ratios obtained using the Manders method tend to under-represent syringyl units, while deconvolution of the same data produce S/G ratios that are consistent with HSQC and py-MBMS methods.
While promising, spectral deconvolution of 13C ssNMR data may not be broadly applicable to all biomass types. Lignin from hardwoods is predominantly S and G with minor H monomeric subunits, p-hydroxybenzoates, ferulates (FA) and p-coumarate (pCA). Therefore, the aromatic region of the 13C NMR spectrum of hardwoods like poplar is reasonably simple making spectral deconvolution straightforward. Corn stover, on the other hand, is known to be rich in S, G and H lignin monomers with an abundance of hydroxycinnamates such as FA and (pCA) levels. The S/G ratios determined from deconvolution of the 148–153 ppm region therefore likely overestimate guaiacyl content since FA and pCA moieties have spectral features near 147 ppm. Similarly, some biomass types have significant representations of non-conventional lignin monomers derived from flavonoids (tricin), hydroxystilbenes, and monolignol acetates [4]. In other words, the spectral deconvolution approach demonstrated here applied to poplar may work to characterize a specific type of biomass (hardwoods are particularly promising) but care must be taken when applying across biomass types. This concept is highlighted in our attempt to deconvolute 13C ssNMR spectrum from corn stover. A ratio between the 153 and 147 ppm regions of 0.7 to 0.8 was observed similar to py-MBMS (0.8), but this does not match the S/G ratio of 1.1 measured from HSQC integrations or 1.3 as measured by thioacidolysis (Fig. 3). When FA content is considered, the S/(G + FA) ratios obtained from HSQC integrations (0.7, data not shown) and solid-state NMR peak-fitting analysis are consistent, confirming that G and FA content cannot be separately quantified in CP-MAS data.
Despite these limitations, solid-state NMR methods are powerful for biomass characterization because they are rich in structural information and data is obtained on samples in their native and unaltered states. For example, in addition to S/G ratios shown here, estimates for cellulose crystallinity index and lignin composition are accessible from the same CP-MAS data [78–80].
In choosing a methodology for studying lignin and particularly if evaluation of S/G ratios is required, the first criteria to consider is size of the sample set. Large sample sets where reliable high-throughput data is required are suitable for py-MBMS or potentially HSQC, but not for thioacidolysis or ssNMR methods. Additionally, if information beyond S/G ratios are required, HSQC spectra can provide bond linkage information as well as other aromatic moieties present in lignin which becomes important for grassy species that are high in coumarates, ferulates and even H-type (p-hydroxyphenyl) lignin. Additionally, py-MBMS could also provide lignin content estimates or thioacidolysis total monomer yields may also be suitable methods if those metrics are needed, although it should be noted that thioacidolysis has the distinct limitation of only cleaving β-O-4 bonds and data will be biased accordingly. Py-MBMS should only compare lignin content of similar biomass types as well and comparing S/G across biomass types by py-MBMS may need to be interpreted with caution as well. If amount of sample is limited, then non-destructive methods may offer the best alternative, although care needs be taken when employing spectral deconvolution to estimate S/G ratios and it is recommended for use only by those with experience in spectral deconvolution. Both HSQC and ssNMR provide a look at intact cell walls, and while it could be argued that the ball-milling required for HSQC may affect cell wall structure, it is likely negligible so long as overheating is prevented and samples being compared are milled consistently. Table 6 summarizes the main considerations for each methodology reported here.
Table 6
Benefits and shortcomings associated with methods used for S/G analysis reported in this work.
Method | Benefit | Shortcoming |
Thioacidolysis | Small sample size, high reproducibility, wealth of historical data | Potential bias for monomers released by β-O-4 linkages not being representative, sensitive to other components in biomass impacting reaction, laborious sample preparation |
Py-MBMS | Small sample size, rapid analysis, high reproducibility, multiple cell wall phenotype measurements possible | Destructive, requires comparison within species, potential bias for monomers released by thermally labile linkages, semi-quantitative |
1H-13C HSQC | Representative of whole cell wall, multiple cell wall phenotype measurements possible particularly including lignin linkage information | Semi-quantitative, large sample size requirement, long analysis time |
ssNMR Manders | Non-Destructive | Underestimates contribution from S lignin monomers |
ssNMR Deconvolution | Non-destructive, Representative of whole cell wall, multiple cell wall phenotype measurements possible | Low throughput, sensitive to incorrect initial peak-fitting parameters, not appropriate for grass species, need other a priori data |