3.1 Initial examination of reported 13C chemical shifts
To obtain 13C chemical shifts for the Iα and Iβ allomorphs of native cellulose, the detailed work of Kono and co-workers was first consulted (Kono et al. 2002, 2003; Kono and Numata 2006). By employing a combination of selective 13C-labelling (Kono et al. 2002), through-bond (Kono et al. 2003) and through-space (Kono and Numata 2006) 2D NMR correlation experiments, full peak assignments of the 13C spectra for cellulose Iα and Iβ were reported. However, as the tables and figures in these papers were closely examined, a number of discrepancies in the reported 13C chemical shifts became evident to us.
To illustrate these discrepancies, two of the key 1D spectra in Kono et al. (2003) were carefully digitized and compared to the chemical shifts reported in the accompanying table in the same paper. Figure 1 displays digitized versions of the Iα and Iβ sub-spectra reported in Figs. 2b and 3b of Kono et al. (2003) which were generated by the authors by taking linear combinations of spectra of Iα-rich Cladaphora and Iβ-rich tunicate celluloses. The chemical shifts at the various maxima in the digitized spectra are indicated above the peaks. In addition, the chemical shifts reported by the authors in Table 1 of Kono et al. (2003) are plotted as stick plots in Fig. 1 and coded with solid and dashed lines according to the glucose unit they belong to, as established by the 2D INADEQUATE correlation spectra presented in Kono et al. (2003).
Upon careful examination of Fig. 1, there are two things that stand out from a comparison of the digitized spectra and the reported chemical shifts. The first is that the chemical shifts for the C2 and C5 signals of cellulose Iα reported by the authors in Table 1 as 70.1 ppm clearly do not agree with the experimental spectrum. The most likely explanation is that this is a typographical error; the chemical shifts probably should be reported as 71.0 ppm. It is worth noting that this same error shows up in a subsequent paper by Kono and Numata (2006) and also in the paper by Wang et al. (2016) which uses these chemical shifts to propose the existence of new cellulose phases in plant primary cell walls. The second thing that stands out is that while most of the reported chemical shifts are close to the peaks in the 1D spectra, they do not agree exactly. Most of the reported chemical shifts are within 0.2 ppm of the corresponding peak maxima, but there are some that have greater differences. For example, the reported chemical shift for the downfield C1 signal of cellulose Iβ at 106.1 ppm deviates by 0.4 ppm from the actual peak in the spectrum at 105.7 ppm. The reported chemical shift for the upfield C3 signal of cellulose Iα at 73.9 ppm deviates by 0.5 ppm from the actual peak at 74.4 ppm. The likely reason for these discrepancies is that the authors reported the chemical shifts estimated from the 2D INADEQUATE spectra, rather than the chemical shifts observed in the 1D spectra. It should be noted that at the relatively low magnetic field that the INADEQUATE spectrum was collected at (7.05 T), the observed correlation peaks have a complicated line shape due to the 13C-13C J-couplings that are being enhanced in this through-bond double-quantum correlation experiment. For this reason, and the low signal-to-noise ratio of the 2D experiment, obtaining the exact chemical shifts from such 2D spectra can be difficult.
A third issue with these reported 13C chemical shifts is how the chemical shift scale was referenced. The authors state that the chemical shift scale was established by setting the carbonyl signal of glycine to 176.03 ppm. It is not clear where this particular reference value comes from, but others have reported the chemical shift of the α polymorph of glycine (with respect to liquid TMS at 0 ppm) to be 176.50 ppm (Potrzebowski et al. 1998) or 176.45 ppm at 296 K with a degree of variability with temperature (Taylor and Dybowski 2008). As will be shown below, inconsistent chemical shift referencing is found throughout the literature on solid-state 13C NMR of cellulose and this issue will be discussed in much greater detail below.
This work by Kono and co-workers in assigning the peaks in the 13C spectra of cellulose Iα and Iβ using 13C-labelling, and through-bond and through-space 2D NMR correlation experiments has been a tremendously important advancement in understanding the solid-state 13C NMR spectra of cellulose. However, unfortunately there are discrepancies in the reported 13C chemical shift values that require correction.
Next, other papers reporting solid-state 13C NMR spectra of cellulose were consulted. However, it became apparent that there is little consistency in the reported 13C chemical shifts in the cellulose NMR literature, exasperating the problems described above. To illustrate this variability, four additional papers that reported solid-state 13C NMR spectra of Iα-rich and Iβ-rich spectra were selected (VanderHart and Atalla 1984; Yamamoto and Horii 1993; Larsson et al. 1999; Kono et al. 2002) and these spectra were carefully digitized (see Table 1 for details). Figure 2a displays digitized spectra of Iα-rich spectra (from Valonia, Cladaphora, or Iα sub-spectra generated by the authors by linear combination). The highlighted region shows that the C1 chemical shift varies over a range of almost 2 ppm, from 105.1 ppm to 106.9 ppm. Other papers report this C1 signal as low as 104.9 ppm (Erata et al. 1997; Sternberg et al. 2003). Figure 2b displays the digitized spectra of Iβ-rich spectra (from tunicate cellulose, annealed Valonia, or Iβ sub-spectra). The highlighted region shows the downfield C1 signal varying over a range of almost 2 ppm from 105.7 ppm to 107.6 ppm.
To move forward with any additional analysis of cellulose structures, for example quantum chemical calculations of 13C chemical shifts, it is necessary to resolve these discrepancies found throughout the cellulose solid-state NMR scientific literature and determine what the correct 13C chemical shifts actually are.
Table 1
Summary of the digitized cellulose spectra presented in Figs. 2 and 3.
Spectrum | Figure | Cellulose type | C1 shift (ppm) | Reference material | Reference shift (ppm) | Correction (ppm) |
Iα-rich cellulose spectra | | | | | | |
VanderHart and Atalla 1984 | Figure 5b | Iα sub-spectrum | 106.3 | LPEa | 33.63 | –0.79 |
Yamamoto and Horii 1993 | Figure 1a | Valonia | 105.6 | LPEa | 32.89 | –0.05 |
Larsson et al. 1999 | Figure 1b | Cladaphora | 105.1 | glycine | 176.03 | + 0.38 |
Kono et al. 2002 | Figure 1a | Cladaphora | 106.9 | glycineb | 176.03b | –1.40 |
Kono et al. 2003 | Figure 2b | Iα sub-spectrum | 105.2 | glycine | 176.03 | + 0.38 |
Iβ-rich cellulose spectra | | | | | | |
VanderHart and Atalla 1984 | Figure 5c | Iβ sub-spectrum | 107.0 | LPEa | 33.63 | –0.79 |
Yamamoto and Horii 1993 | Figure 1d | annealed Valonia | 106.2 | LPEa | 32.89 | –0.05 |
Larsson et al. 1999 | Figure 1d | Tunicate | 105.7 | glycine | 176.03 | + 0.38 |
Kono et al. 2002 | Figure 2d | Tunicate | 107.6 | glycineb | 176.03b | –1.40 |
Kono et al. 2003 | Figure 3b | Iβ sub-spectrum | 105.7 | glycine | 176.03 | + 0.38 |
a LPE = linear polyethylene |
b Paper reports setting carbonyl signal of glycine to 176.03 ppm, but a mistake must have occurred (see text) |
3.2 Correcting reported 13C chemical shifts
The primary reason for the variability in the spectra presented in Fig. 2 is that different secondary chemical shift reference compounds have been used by various authors (see Table 1). However, not only is there a variety of reference compounds used in the literature on solid-state 13C NMR of cellulose, the values used to set the chemical shift axis are not always consistent nor correct. In the section that follows, it is explained how these spectra can be corrected and put on the same correct chemical shift scale (with respect to liquid TMS at 0 ppm).
In the pioneering work by Vanderhart and Atalla (Atalla and Vanderhart 1984; VanderHart and Atalla 1984), in which it was first demonstrated that most native celluloses are composites of two crystalline phases Iα and Iβ, the chemical shift reference material was a sample of crystalline linear polyethylene (LPE) added to the sample and set to 33.63 ppm. These original Iα and Iβ sub-spectra, collected at a magnetic field of 4.7 T, were carefully digitized and are reproduced here in the top of Fig. 2a and Fig. 2b, respectively. Vanderhart later reported the existence of a field-dependence of 13C chemical shifts due to a second-order effect of 1H-13C dipolar interactions (VanderHart 1986). This effect is small, but most pronounced for CH2 carbons at low magnetic fields. The field-dependence of the LPE 13C chemical shift was studied and it was shown that it changed from 33.63 ppm at 1.4 T to 32.89 ppm at 4.7 T (and plateauing at 32.82 ppm at higher magnetic fields). Since the cellulose spectra reported by Vanderhart and Atalla and reproduced here were collected at a higher field (4.7 T), but referenced with the LPE chemical shift value measured at low field (1.4 T), it is necessary to correct for this field-dependent effect by adjusting the chemical shifts by − 0.74 ppm. In addition, one other small adjustment needs to be made. In the study of the field-dependence of 13C chemical shifts (VanderHart 1986), the chemical shifts were referenced with respect to the CH2 signal of adamantane at 29.50 ppm with respect to liquid TMS. Later work (Morcombe and Zilm 2003) has shown that the CH2 signal of adamantane is actually at 29.45 ppm from liquid TMS, requiring an additional adjustment of − 0.05 ppm. Therefore, the overall correction for these spectra is − 0.79 ppm in order to give a chemical shift axis with respect to liquid TMS at 0 ppm.
In a series of spectra which showed that it was possible to convert Iα-rich Valonia cellulose into a Iβ rich form by annealing the sample at high temperatures, Yamamoto and Horii (1993) used LPE as a secondary reference compound by correctly setting its chemical shift to 32.89 ppm (spectra were obtained at 4.7 T). This means only a correction of − 0.05 ppm is required (the difference in the adamantane chemical shifts described above) to put these spectra on a chemical shift axis with respect to liquid TMS at 0 ppm.
In numerous papers reporting solid-state 13C NMR spectra of cellulose, including three from which spectra in Fig. 2 were digitized (Larsson et al. 1999; Kono et al. 2002, 2003), the secondary reference material used was glycine with its carbonyl signal set to 176.03 ppm. It is not clear where this chemical shift value comes from, but the earliest paper reporting a 13C NMR spectrum of cellulose using this value we could find is a paper by Lennholm et al. (1994). As mentioned above, other authors have reported the chemical shift of the α polymorph of glycine (with respect to liquid TMS) to be 176.50 ppm (Potrzebowski et al. 1998) or 176.45 ppm at 296 K with a degree of variability with temperature (Taylor and Dybowski 2008). To investigate this for ourselves, we collected 13C CP MAS spectra of α-glycine at two fields (4.7 T and 21.1 T) immediately after collecting a 13C CP MAS spectrum of adamantane and setting its CH signal to 38.48 ppm from liquid TMS, as reported by Morcombe and Zilm (2003). The observed chemical shift for the carbonyl signal of α-glycine at both fields was 176.41 ppm. Therefore, in order to put spectra which have employed a chemical shift reference value of 176.03 ppm for glycine onto a chemical shift axis with respect to liquid TMS at 0 ppm it is necessary to adjust the chemical shifts by + 0.38 ppm.
It is worth making mention of the spectra reported by Kono et al. (2002) in their 13C-labelling study. These spectra are displayed second from the bottom in Fig. 2a and 2b. It is stated in this paper that the chemical shift axis was set by setting the carbonyl signal of glycine to 176.03 ppm. However, there must have been a mistake in the referencing since the reported chemical shifts deviate by about 1.8 ppm from the other two papers that also report using glycine as the chemical shift reference material (see Table 1). Confusingly, when referring to these chemical shifts in their next paper, Kono et al. (2003) report very different values in their table than what was reported in Kono et al. (2002). We found that it was necessary to adjust the chemical shifts by − 1.40 ppm to bring these spectra in line with all of the others.
Applying the corrections described above (summarized in the last column of Table 1) to each spectrum in Fig. 2 so that all spectra are on the same chemical shift axis referenced to liquid TMS at 0 ppm yields a set of Iα-rich and Iβ-rich spectra that are excellent agreement with each other, as shown in Fig. 3. The vertical lines are the chemical shifts determined from the Iα and Iβ sub-spectra reported by Kono et al. (2003) after the chemical shift axes were corrected (bottom spectra in Fig. 3a and Fig. 3b). These final peak assignments and chemical shifts are listed in Table 2 and will be elaborated on in the following section. For now, these vertical lines are helpful in guiding the eye and showing how well these spectra agree with each other now that they are all on the same chemical shift scale.
Table 2
Correctly referenced 13C chemical shifts (in ppm from liquid TMS at 0 ppm)a for the glucose units in cellulose Iα and Iβ.
| | C1 | C2 | C3 | C4 | C5 | C6 |
Cellulose Iα | | | | | | | |
glucose unit 1 | (solid orange lines) | 105.48 | 71.24 | 75.11 | 90.29 | 71.34 | 65.83 |
glucose unit 2 | (dashed red lines) | 105.58 | 72.22 | 74.63 | 89.42 | 73.10 | 65.73 |
Cellulose Iβ | | | | | | | |
glucose unit 1 | (solid blue lines) | 106.12 | 71.74 | 75.28 | 88.36 | 71.43 | 65.99 |
glucose unit 2 | (dashed green lines) | 104.39 | 71.74 | 74.44 | 89.18 | 72.85 | 65.22 |
a to convert chemical shifts to the IUPAC-recommended chemical shift reference of dilute (< 1%) tetramethylsilane (TMS) in deuterated chloroform (CDCl3), the values in this table must be adjusted by -0.71 ppm.
3.3 Obtaining definitive 13C chemical shifts and assignments
After correcting each of these spectra and putting them on the same chemical shift scale with respect to TMS at 0 ppm, we returned to the Iα and Iβ sub-spectra reported by Kono et al. (2003) that were initially presented in Fig. 1. With these spectra now properly referenced to TMS at 0 ppm, the cellulose Iα and Iβ spectra were fit with model spectra consisting of two Lorentizan peaks for each carbon type, with all the peak areas constrained to be equal to each other reflecting the equal populations of each site in the crystal structures. For cellulose Iα, the fitting initially resulted with the C1, C6, and one of the C2/C5 pairs with identical chemical shifts. A detailed analysis of a 2D correlation spectrum of bacterial cellulose suggested small differences of about 0.1 ppm in these pairs of signals (see below) and these small differences were incorporated into the fit presented here. These model spectra for cellulose Iα and Iβ, presented as the lower grey spectra in Fig. 4a and Fig. 4b respectively, are in excellent agreement with the digitized and corrected experimental spectra. The resulting chemical shifts and their assignments are presented as stick plots. These chemical shifts are also listed in Table 2 and are the same chemical shifts used to generate the vertical lines in Fig. 3 that show the excellent agreement between a variety of Iα-rich and Iβ-rich spectra taken from the literature, once the chemical shift axes were corrected.
For the cellulose Iβ spectrum, the assignment of the carbon signals to the two glucose units proposed by Kono et al. (2003) based on the 2D INADEQUATE spectrum was retained. Here we use solid blue lines with bold font and dashed green lines with italic font to distinguish the signals from the two glucose units. It is worth commenting that there is actually some ambiguity of the assignment of the C1 signals based on the 2D INADEQUATE spectrum alone since the C2 signals have identical chemical shifts. However, Kono and Numata (2006) later reported 2D radio-frequency dipolar recoupling (RFDR) correlation spectra of cellulose Iβ that support this assignment. By probing longer range correlations between carbons, they observed C1-C3 correlations (within a glucose unit) and C1-C4 correlations (across the glycosidic C1-O-C4 linkage) that are consistent with this assignment of the C1 signals.
For the cellulose Iα spectrum, a close comparison of the peak assignments for cellulose Iα presented in Figs. 1a and 4a reveals that we have swapped the assignment of the C2 signals from one glucose unit to the other compared to what Kono et al. (2003) originally proposed based on their analysis of a 2D INADEQUATE spectrum. The evidence for this change comes from a careful analysis of a 2D NMR correlation experiment that we performed on a Iα-rich bacterial cellulose sample, as shown in Fig. 5 and discussed below.
A 13C-enriched bacterial cellulose sample was biosynthesized using 25% U-13C6-glucose as the carbon source. The 1D 13C CP MAS NMR spectrum (Fig. 5a) reveals the presence of multiple forms of cellulose, but that the dominant form is the Iα phase. The spectrum was referenced with respect to TMS at 0 ppm, using adamantane as a secondary reference by setting its CH signal to 38.48 ppm. It was very satisfying that the chemical shifts (Table 2) determined by fitting the digitized and corrected Iα sub-spectrum (Fig. 4a) gave excellent agreement with the main peaks in this independently collected spectrum of Iα-rich bacterial cellulose, as shown by the chemical shift stick plot in Fig. 5a.
With the modest 13C enrichment of the sample, it was relatively straightforward to collect a 2D double-quantum correlation spectrum to probe the carbon-carbon connectivities in the material (Fig. 5b). To obtain this spectrum, a through-space dipolar recoupling pulse sequence was employed with a sufficiently short dipolar recoupling time to probe only nearest-neighbor 13C-13C correlations. While this bacterial cellulose sample has multiple cellulose forms present, the 2D spectrum is dominated by correlations arising from the Iα phase.
The chemical shifts determined by fitting the digitized and corrected Iα sub-spectrum (see Fig. 4 and Table 2) were used to generated the transparent orange and red correlations superimposed on the 2D spectrum. Pairs of signals arising from carbons in close proximity (nearest-neighbor carbons in this spectrum) appear as correlations at their respective isotropic chemical shifts in the direct dimension and at the sum of their chemical shifts in the indirect double quantum dimension. The zoomed regions in Fig. 5c show there is excellent agreement between the chemical shifts and the observed correlations. The fact that the proposed chemical shifts agree with both the individual chemical shifts in the direct dimension and at sums involving other correlated peaks in the indirect double-quantum dimension provides a very high degree of confidence in these proposed chemical shifts and their assignments.
Initially, the assignments of the carbon signals to the two glucose units in cellulose Iα proposed by Kono et al. (2003) were used. For nearly all the correlations, the agreement with the 2D DQ correlation spectrum in Fig. 5b is excellent, except for the C3-C2 correlation as shown in the inset in Fig. 5d. We found that by exchanging the glucose units to which the C2 signals were assigned gave much better agreement with the C3-C2 correlation, as shown on the right in Fig. 5c. Swapping the C2 signals has no significant impact on the C1-C2 correlation since the C1 chemical shifts are nearly identical.
In addition, a close inspection of the shapes of the correlations involving C1 and C6, as well as the C2/C5 pairs of signals from glucose unit 1 suggests that these pairs of C1, C6, and C2/C5 chemical shifts may be slightly different from each other. Our final peak assignment includes these signals differing by 0.1 ppm, rather than being identical to each other. These chemical shifts were fixed to these values when the Iα sub-spectrum in Fig. 4a was fit.
Through this process of digitization, chemical shift axis correction, spectrum fitting, and careful analysis of this 2D correlation spectra, we have arrived at we believe is a definitive set of 13C chemical shifts for all signals in the 13C NMR spectra of cellulose Iα and Iβ, as well as the correct assignment of these signals to the two glucose units in each of these structures.
Given these results, a word of caution should be made about the proposal by Wang et al. (2016) that there may be up to 7 additional phases of cellulose found in plant primary cell walls, particularly the presence of 5 new interior crystalline phases. These new phases were proposed based on differences in their observed chemical shifts and the chemical shifts reported in Table 1 of Kono et al. (2003) which we have demonstrated here not to be entirely correct for a variety of reasons (a typographical error, a different secondary chemical shift reference, and differences between chemical shifts in the 1D and 2D spectra). It would be interesting to see how their analysis changes if the set of 13C chemical shifts presented here was used. A similar caution should be extended as well to the “RMSD map” analysis reported by Kirui et al. (2019) in which the chemical shifts of Kono et al. (2003) were presumably used for the Iα and Iβ phases.
It is worth a final comment about the 13C chemical shifts. While it remains widespread practice in the solid-state NMR community to report 13C NMR spectra on a chemical shift scale with respect to neat liquid TMS at 0 ppm, usually by setting the CH signal of adamantane to 38.48 ppm (Morcombe and Zilm 2003), a working group from the International Union of Pure and Applied Chemistry (IUPAC) has actually recommended using a dilute solution (< 1%) of TMS in deuterated chloroform (CDCl3) as a universal chemical shift reference for all nuclei (Harris et al. 2008). The 13C chemical shifts of neat TMS and a dilute solution of TMS in CDCl3 differ by 0.71 ppm and it is recommended to set the CH signal in adamantane to 37.77 ppm (Harris et al. 2008; Hoffman 2022). Therefore, in order to convert these reported 13C chemical shifts for cellulose to the IUPAC-recommended chemical shift scale with respect to a dilute solution of TMS in CDCl3 at 0 ppm, it would be necessary to adjust the chemical shifts by -0.71 ppm, as mentioned in the footnote in Table 2.