Thioacidolysis
Thioacidolysis yields of aromatic lignin units for each of the biomass samples are presented in Table 1 (also Supplementary Table 1 with other S/G values determined by other methods). The Pearson correlation coefficient between the total yield of thioacidolysis-released aromatic lignin units on a g-1 of biomass basis and the yield of aromatic lignin units reported g-1 of estimated lignin content (based on py-MBMS analysis) was 0.93; therefore, comparison of yields on g-1 of estimated lignin content or a g-1 of biomass would be similar and are subsequently discussed on a g-1 of biomass basis as the lignin content was an estimate based on py-MBMS analysis. While the H unit content determined by thioacidolysis was consistent among the set of poplar samples, the S and G unit content varied widely, both in S/G ratio and total aromatic lignin unit yields. The S/G ratio determined by thioacidolysis did not correlate with the total yield of aromatic units (Pearson correlation = -0.08, Supplementary Table 2). However, the yield of S units was the primary driver in the total yield of thioethylated aromatic lignin units for data both within the set of poplar samples (Pearson correlation of S, total units = 0.92, Pearson correlation of G, total units = 0.62) and including the corn stover (Pearson correlation of S, total units = 0.94, Pearson correlation of G, total units = 0.68). Thioacidolysis of pine produced only G and H-derived units, similar to previously reported values [18]. The corn stover sample yielded lower total aromatic lignin units from thioacidolysis than most of the hardwoods and softwood samples, which is typical for grasses [17, 18].
Table 1. Aromatic lignin unit content of biomass determined by thioacidolysis (average of n=2 samples, not all samples were analyzed by thioacidolysis and are denoted N/A).
Biomass ID
|
H µmol/g biomass
|
S µmol/g biomass
|
G µmol/g biomass
|
S/G
|
Sum µmol/g biomass
|
Sum µmol/g lignin content*
|
NIST 8493 Pine
|
6.7
|
0.0
|
205.7
|
0.0
|
217.7
|
818.4
|
FCIC Corn Stover
|
4.8
|
47.5
|
37.9
|
1.3
|
97.2
|
709.5
|
BESC-004 Poplar
|
5.5
|
218.5
|
105.6
|
2.1
|
329.6
|
1277.5
|
BESC-021 Poplar
|
5.5
|
111.3
|
87.5
|
1.3
|
204.3
|
908.0
|
BESC-036 Poplar
|
5.4
|
207.3
|
69.4
|
3.0
|
282.0
|
1155.7
|
BESC-075 Poplar
|
4.5
|
78.1
|
22.3
|
3.5
|
104.9
|
476.8
|
BESC-095 Poplar
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-096 Poplar
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-140 Poplar
|
5.0
|
156.3
|
40.7
|
4.0
|
202.0
|
834.7
|
BESC-169 Poplar
|
5.1
|
144.8
|
85.5
|
1.7
|
235.3
|
1089.4
|
BESC-173 Poplar
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-182 Poplar
|
5.1
|
185.3
|
78.1
|
2.4
|
268.5
|
1214.9
|
BESC-217 Poplar
|
5.1
|
147.0
|
57.3
|
2.6
|
209.4
|
887.3
|
BESC-219 Poplar
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-255 Poplar
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-282 Poplar
|
5.0
|
119.5
|
87.1
|
1.4
|
211.5
|
1026.7
|
BESC-322 Poplar
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-334 Poplar
|
5.0
|
198.1
|
68.4
|
2.9
|
271.6
|
N/A
|
BESC-388 Poplar
|
4.8
|
132.2
|
104.7
|
1.3
|
241.8
|
1129.9
|
BESC-841 Poplar
|
5.1
|
231.4
|
68.3
|
3.4
|
304.8
|
1354.7
|
BESC-853 Poplar
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-863 Poplar
|
4.8
|
138.6
|
62.9
|
2.2
|
206.3
|
916.9
|
BESC-883 Poplar
|
4.8
|
189.1
|
60.4
|
3.1
|
254.3
|
1077.5
|
*Lignin content estimate based on py-MBMS analysis
Py-MBMS
Py-MBMS analysis of the biomass samples present in sufficient quantity (not all samples were analyzed due to low mass availability) produced spectra consisting of ions derived from S, G, and H lignin units bound by various types of linkages. Lignin contents were determined based on methods described previously [29, 31, 32] in order to estimate Klason lignin content (wt %) using mean-normalized spectra to remove mass-dependent variation; S/G ratios were determined using unique (minimal-overlapping) ions of known origin that produced S/G ratios consistent with other methods in the literature [31, 70]. Traditional S/G ratios determined by py-MBMS are assumed as being derived from S, G, and H units where this method was established based on the results from NMR and thioacidolysis, but theoretically could include additional or alternative ions, particularly as some of the ions chosen to determine S/G do potentially originate from multiple sources to varying degrees.
Lignin content estimates of the poplar samples by py-MBMS ranged from 20.6 to 25.8 wt% lignin (Table 2). Traditional S/G ratios determined for the poplar samples using py-MBMS ranged from 1.3 to 2.2. Focusing on ions between 50-250 due to the nature of ions outside of that range originating primarily from noise and overlapping sources, the variance of the spectra was highest for m/z 138 (G lignin), 151 (G lignin, this ion may also originate from ferulate), 165 (S lignin), and 181 (S lignin), where () indicates origin of ion. The lignin content and S/G based on py-MBMS data were weakly correlated (Pearson correlation = 0.54) within the poplar samples. Lignin content as determined by py-MBMS did not strongly correlate with thioacidolysis yields for the entire poplar set either (Pearson correlation = 0.45, Supplementary Table 2). However, the majority of the poplar lignin estimates from py-MBMS did appear to correlate well with aromatic peak abundances from the peak-fit NMR data (values in Supplementary Table 1). The lignin content of the NIST 8493 Monterey pine was provided by the supplier based on Klason results to be 26.6 wt% lignin, and the lignin content of the FCIC corn stover was estimated as 13.7 wt% relative to a corn stover sample of known Klason content using py-MBMS spectra. Table 2 provides characterization data of the biomass samples based on py-MBMS analysis. The S/G ratio determined for the NIST pine was 0.2 since non-zero values of the ions otherwise derived from S units were observed in the spectra, but these values could not be differentiated from noise and fragment ions and because pine does not produce S lignin, the S/G for NIST pine was assigned to 0. Because the type and amount of noise and/or fragmentation contributing to the abundance of these ions is not known or likely to be consistent for each of the biomass types, this value was not adjusted across the other samples.
Table 2. Py-MBMS characterization of lignin content and S/G ratios in select biomass samples (not all samples were analyzable by py-MBMS and are denoted N/A).
Biomass ID
|
S/G
|
Lignin content
|
NIST 8493 Pine
|
0.0*
|
26.6
|
FCIC Corn Stover
|
0.8
|
13.7
|
BESC-004 Poplar
|
1.9
|
25.8
|
BESC-021 Poplar
|
1.3
|
22.5
|
BESC-036 Poplar
|
2.1
|
24.4
|
BESC-075 Poplar
|
2.0
|
22.0
|
BESC-095 Poplar
|
N/A
|
N/A
|
BESC-096 Poplar
|
N/A
|
N/A
|
BESC-140 Poplar
|
2.0
|
24.2
|
BESC-169 Poplar
|
1.6
|
21.6
|
BESC-173 Poplar
|
N/A
|
N/A
|
BESC-182 Poplar
|
1.9
|
22.1
|
BESC-217 Poplar
|
2.0
|
23.6
|
BESC-219 Poplar
|
N/A
|
N/A
|
BESC-255 Poplar
|
N/A
|
N/A
|
BESC-282 Poplar
|
1.4
|
20.6
|
BESC-322 Poplar
|
N/A
|
N/A
|
BESC-334 Poplar
|
N/A
|
N/A
|
BESC-388 Poplar
|
1.4
|
21.4
|
BESC-841 Poplar
|
2.2
|
22.5
|
BESC-853 Poplar
|
N/A
|
N/A
|
BESC-863 Poplar
|
1.8
|
22.5
|
BESC-883 Poplar
|
2.1
|
23.6
|
*adjusted value based on S-derived lignin ion intensities low value
Gel state NMR (HSQC)
Gel-state HSQCs were collected on all poplar samples, as well as corn stover and pine, according to Mansfield et al. [71]. Generally, S/G ratios were consistent over triplicate runs, although the percent error ranged from as low as 2% to as high as 16%. It is believed that this range in error is partly due to inconsistencies in how well individual samples formed a gel (visual observation) and could also possibly come from the milling process. Some samples demonstrated better “gelling” after being heated to 40°C for data collection, whereas other samples were unaffected (visual observation, data not shown). The S/G ratio of the corn stover and poplar samples ranged from 0.9 to 2.4. The S/G ratios of the poplar samples as determined by HSQC correlated strongly with the S/G ratios as determined by thioacidolysis (although the range for thioacidolysis was broader) and py-MBMS (Pearson correlation, HSQC/thioacidolysis = 0.84, HSQC/py-MBMS = 0.96, Supplementary Table 2).
Lignin unit linkages in the poplar samples as determined by HSQC provided structural information to inform bias in methodologies used to measure S/G ratios. Interestingly, β-O-4 linkages as determined by HSQC did not correlate with thioacidolysis yields (Pearson correlation = 0.11) and did not correlate strongly with lignin content as determined by py-MBMS (Pearson correlation = 0.45), indicating that the total yields of aromatic lignin units as detected from thioacidolysis and by py-MBMS was not solely dependent on the abundance of those linkages in the biomass but is complicated by the other linkages as well. Additionally, the poplar β-O-4 linkages did not strongly correlate with S lignin unit yields from thioacidolysis (Pearson correlation = 0.40) but did for py-MBMS ions typically derived from S lignins (Pearson correlation for most S-derived ions ~0.6, Supplementary Table 2). However, β-O-4 linkages in the poplar set did more strongly correlate with S/G ratio as determined by py-MBMS (Pearson correlation = 0.71), and to a lesser degree with S/G as determined by thioacidolysis (Pearson correlation = 0.70) and HSQC (Pearson correlation = 0.61). There was a general weakly negative correlation of poplar β-O-4 linkages with G units (Pearson correlation, HSQC/thioacidolysis = -0.51, HSQC/G units from py-MBMS ~ -0.5). Table 3 lists the calculated S/G and bond content results of triplicate analyses of poplar, pine, and corn stover samples.
Table 3. HSQC calculated S/G ratios and bond content.
|
Biomass ID
|
S/G Ratio
|
S/G Error
|
β-O-4
|
Error
|
β-β
|
Error
|
β-5
|
Error
|
NIST 8493 Pine
|
N/A
|
N/A
|
57
|
N/A
|
10
|
N/A
|
33
|
N/A
|
FCIC Corn Stover
|
1.1
|
0.20
|
99
|
N/A
|
N/A
|
N/A
|
N/A
|
N/A
|
BESC-004
|
1.6
|
0.15
|
70
|
2.0
|
24
|
1.4
|
6.1
|
1.2
|
BESC-021
|
0.89
|
0.04
|
66
|
2.6
|
25
|
2.4
|
9.4
|
1.5
|
BESC-036
|
1.9
|
0.17
|
74
|
3.8
|
21
|
4.1
|
4.8
|
2.1
|
BESC-075
|
2.1
|
0.05
|
72
|
2.3
|
23
|
2.1
|
4.6
|
0.4
|
BESC-095
|
1.5
|
0.21
|
75
|
2.6
|
20
|
1.2
|
5.2
|
2.4
|
BESC-096
|
2.1
|
0.25
|
76
|
2.3
|
20
|
2
|
3.4
|
0.9
|
BESC-140
|
1.7
|
0.11
|
74
|
4.2
|
22
|
2.8
|
4.3
|
1.5
|
BESC-169
|
1.2
|
0.05
|
73
|
1.9
|
21
|
0.5
|
6.5
|
2.1
|
BESC-173
|
1.1
|
0.13
|
69
|
2.1
|
23
|
1.1
|
8.2
|
1.0
|
BESC-182
|
1.8
|
0.20
|
73
|
4.1
|
21
|
3.7
|
5.9
|
2.3
|
BESC-217
|
1.7
|
0.18
|
72
|
0.5
|
23
|
2.1
|
5.3
|
1.8
|
BESC-219
|
2.3
|
0.13
|
72
|
1.8
|
25
|
3.1
|
2.8
|
1.6
|
BESC-255
|
1.8
|
0.17
|
71
|
3.4
|
22
|
3.0
|
6.2
|
1.0
|
BESC-282
|
1.1
|
0.04
|
62
|
3.7
|
25
|
2.6
|
14
|
4.8
|
BESC-322
|
2.2
|
0.19
|
75
|
2.4
|
22
|
1.9
|
3.2
|
0.8
|
BESC-334
|
2.3
|
0.21
|
69
|
2.3
|
27
|
1.4
|
4.5
|
1.0
|
BESC-388
|
0.94
|
0.01
|
67
|
4.4
|
22
|
2.4
|
11
|
3.1
|
BESC-841
|
2.4
|
0.30
|
73
|
5.7
|
23
|
5.9
|
3.8
|
1.8
|
BESC-853
|
2.4
|
0.33
|
73
|
3.8
|
23
|
1.6
|
3.3
|
2.4
|
BESC-863
|
1.4
|
0.12
|
73
|
4.3
|
23
|
2.9
|
4.1
|
1.7
|
BESC-883
|
2.3
|
0.36
|
76
|
1.8
|
21
|
2.9
|
2.5
|
1.4
|
Solid-State NMR – Manders Subtraction Method
The S/G ratios of poplar samples were calculated from solid-state 13C NMR interrupted decoupling spectra as described by Manders [57]. Table 4 gives the integral values and the S/G ratios calculated using the Manders method for the poplar samples. A low, mid-range, and two high S/G ratio samples (as calculated by py-MBMS) were run in triplicate to determine consistency of the Manders method. Generally, S/G ratios were consistent over triplicate runs, with the percent error ranging from 4% to 7%. Overall, the S/G ratios calculated trend much lower than all the other methods studied thus far and had a significantly lower range of S/G ratios. The S/G ratio of the corn stover and poplar samples ranged from 0.5 to 1.1. The S/G ratios of the poplar samples as determined by ssNMR still did correlate with the S/G ratios as determined by thioacidolysis, py-MBMS or HSQC (Pearson correlation, ssNMR/thioacidolysis = 0.76, ssNMR/py-MBMS = 0.81, ssNMR/HSQC = 0.80).
Table 4. Integration values of interrupted decoupling ssNMR spectra (Manders method)
|
Biomass ID
|
S Integral
|
G Integral
|
Normalized S
|
Normalized G
|
S/G
|
NIST 8493 Pine
|
N/A
|
100.0
|
N/A
|
33.3
|
N/A
|
FCIC Corn Stover
|
37.6
|
62.4
|
9.4
|
20.8
|
0.5
|
BESC-004
|
48.1
|
51.9
|
12.0
|
17.3
|
0.7
|
BESC-021*
|
46.4
|
53.6
|
11.6
|
17.9
|
0.7
|
BESC-036
|
52.3
|
47.7
|
13.1
|
15.9
|
0.8
|
BESC-075
|
53.2
|
46.8
|
13.3
|
15.6
|
0.9
|
BESC-095
|
52.8
|
47.2
|
13.2
|
15.7
|
0.8
|
BESC-096
|
59.2
|
40.8
|
14.8
|
13.6
|
1.1
|
BESC-140
|
54.8
|
45.2
|
13.7
|
15.1
|
0.9
|
BESC-169
|
43.7
|
56.3
|
10.9
|
18.8
|
0.6
|
BESC-173
|
47.5
|
52.5
|
11.9
|
17.5
|
0.7
|
BESC-182*
|
48.8
|
51.2
|
12.2
|
17.1
|
0.7
|
BESC-217
|
54.0
|
46.0
|
13.5
|
15.3
|
0.9
|
BESC-219
|
59.0
|
41.0
|
14.7
|
13.7
|
1.1
|
BESC-255
|
53.4
|
46.6
|
13.4
|
15.5
|
0.9
|
BESC-282
|
43.1
|
56.9
|
10.8
|
19.0
|
0.6
|
BESC-322
|
50.9
|
49.1
|
12.7
|
16.4
|
0.8
|
BESC-334
|
59.8
|
40.2
|
15.0
|
13.4
|
1.1
|
BESC-388*
|
45.7
|
54.3
|
11.4
|
18.1
|
0.6
|
BESC-841
|
54.6
|
45.4
|
13.7
|
15.1
|
0.9
|
BESC-853*
|
55.8
|
44.2
|
14.0
|
14.7
|
0.9
|
BESC-863
|
50.1
|
49.9
|
12.5
|
16.6
|
0.8
|
BESC-883
|
59.7
|
40.3
|
14.9
|
13.4
|
1.1
|
* Average of duplicate or triplicate sample runs
Solid-state NMR – Spectral Deconvolution
Spectral deconvolution (peak-fitting) of 13C CP-MAS spectra in the aromatic domain was performed to estimate the relative abundances of S and G and lignin for twenty-two natural poplar variants and one corn stover sample. Peak-fitting 13C solid-state NMR data to quantitatively understand local molecular structure of biopolymers is a widely used practice, with many applications to biomass and other heterogeneous polymers [67, 72, 73]. Solid-state NMR profiles are generally deconvoluted into pseudo-Voight lineshapes, meaning a weighted sum of Gaussian and Lorentzian profiles are used for a single resonance. This means any particular 13C NMR resonance has variables in peak position, amplitude, width (Full Width Half Max, or FWHM), and finally peak shape (Gaussian vs Lorentzian weighting factor). When deconvoluting overlapping resonances, inaccurate initial estimates of peak position, linewidth and peak shape might result in unreliable and inaccurate fits.
To identify acceptable initial fitting conditions prior to spectral deconvolution, we collected 2D 13C-13C through-space Dipolar Assisted Rotational Resonance (DARR) spectra on model 13C-enriched woody biomass (Figure 1). Precise 13C chemical shifts were extracted from inspection of off-diagonal cross-peaks, and reasonable linewidth estimates were obtained from analysis of resolved cross-peaks. S-lignin peaks were identified from S-lignin rich 13C-enriched hybrid poplar woody stems (Figure 1a), while G-lignin shifts were extracted from 13C-enriched Monterey pine since softwood biomass is entirely G-lignin (Figure 1b). We note that two signals centered at 146.5 and 148.5 could be identified in the 2D spectrum, which we use to represent G3,4 moieties generally. Since guaiacyl units lack a methoxy group at the ring-5 position and are therefore subject to carbon-carbon and carbon-oxygen condensation, a broader distribution of chemical environments is expected for G units compared to the more symmetric S units. Minor G-lignin chemical shifts in the poplar biomass were consistent with G-lignin signals in the Pine sample. Peak shapes were initially set to mixed 90% Gaussian 10% Lorentzian component because sample heterogeneity will impart a Gaussian distribution of Lorentzian-like signals. However, this weighting factor was varied systematically (from 9:1 to 1:9) to help identify fitting errors. To improve these initial starting parameters, a single CP-MAS spectrum representative of the full dataset was deconvoluted such that chemical shifts were only allowed to perturb by 0.1 ppm about the shifts identified from the 2D data, and peak widths and lineshapes were allowed minor deviation from initial guesses to obtain optimal fits. Finally, with initial chemical shifts, peak widths and peak shapes all carefully estimated, batch-fitting of the entire dataset was accomplished using Python code (fitting performed with lmfit module) in which only peak amplitude was allowed to vary for each spectrum whereas peak position, Gaussian/Lorentzian ratios and FWHM were locked for all samples. Representations of resulting spectral deconvolutions for High-S (BESC-096) and Low-S (BESC-021) lignin natural poplar variants are shown in Figure 2 (fits for all samples are provided in Supplementary Materials). S/G ratios arise directly from the relative deconvoluted peak areas of the S3,5 to G3,4 signals.
When applying this fitting strategy, it became clear that small variations in the initial starting conditions had minor effects on the final observable, namely S/G ratios based on the 153/148 ppm peak areas. Since estimation of initial fitting parameters were obtained with a manual process and are therefore subject to some researcher bias, the above procedure was repeated several times. Table 5 lists the average S/G ratios obtained from repeated batch-fitting of the entire poplar and corn stover dataset using a range of different starting conditions.
Table 5. S/G ratios obtained from peak fitting for deconvolution of ssNMR (n=7 fitting iterations).
Biomass ID
|
Average S/G
|
Std Dev
|
FCIC Corn Stover
|
0.74*
|
0.04
|
BESC-004
|
1.6
|
0.1
|
BESC-021
|
1.2
|
0.1
|
BESC-036
|
2.0
|
0.2
|
BESC-075
|
2.1
|
0.2
|
BESC-095
|
1.7
|
0.1
|
BESC-096
|
2.0
|
0.2
|
BESC-140
|
1.9
|
0.2
|
BESC-169
|
1.4
|
0.1
|
BESC-173
|
1.4
|
0.1
|
BESC-182
|
1.7
|
0.1
|
BESC-217
|
1.9
|
0.1
|
BESC-219
|
2.1
|
0.2
|
BESC-255
|
2.1
|
0.2
|
BESC-282
|
1.3
|
0.1
|
BESC-322
|
2.2
|
0.2
|
BESC-334
|
2.2
|
0.2
|
BESC-388
|
1.4
|
0.1
|
BESC-841
|
2.1
|
0.2
|
BESC-853
|
2.1
|
0.2
|
BESC-863
|
1.6
|
0.1
|
BESC-883
|
2.2
|
0.2
|
* Ratio measured = S/(G+FA)