Analytical Approaches to Improved Characterization of Substitution in

For cloud point analysis, HPC samples (∼250 mg) were dissolved in 10 mM NaCl to ..... The two HPC samples from different suppliers were shown to dif...
0 downloads 0 Views 99KB Size
Anal. Chem. 2003, 75, 6077-6083

Analytical Approaches to Improved Characterization of Substitution in Hydroxypropyl Cellulose Sara Richardson,* Thomas Andersson, Gunnar Brinkmalm, and Bengt Wittgren

AstraZeneca R&D Mo¨lndal, SE-431 83 Mo¨lndal, Sweden

Chemical characterization of cellulose derivatives is of high importance as it provides information about the often inhomogeneous substitution that may seriously affect the properties of these polymers in various applications. A detailed mapping of the chemical structure of these derivatives requires several advanced techniques to be employed. In this study, the average substitution and the substitution heterogeneity in two hydroxypropyl cellulose (HPC) samples from different suppliers were studied by means of NMR spectroscopy, MALDI-TOF MS, and HPAEC-PAD. 1H and 13C NMR provided information on the molar substitution, a parameter that could be analyzed by MALDI-TOF MS as well. In addition, the latter technique was used for determination of the distribution of the number of hydroxypropyl groups per glucose unit present in the two polymers. The heterogeneity of the substitution was studied by determining the amount of unsubstituted glucose units in the HPC samples, which was accomplished by HPAEC-PAD analysis. The results obtained suggest that the two HPC samples differ in both hydroxypropoxy content and distribution of the hydroxypropyl groups. Further, the benefits and importance of employing several analytical methods when investigating the cellulose ether substitution are demonstrated, as each method provides different kinds of information on the chemical content. Chemical modification of cellulose to produce cellulose ethers has a tradition since the 1910s and is still of great interest in the development of new polymeric materials. One of the important commercial cellulose ethers is hydroxypropyl cellulose (HPC), which is a water-soluble, nontoxic polymer. HPC is used in a wide range of industrial applications, due to its ability to serve as, for example, colloidal stabilizer, emulsifier, and coating agent in ceramics, paint, paper, or textile.1,2 HPC has also found extensive application in oral and topical pharmaceutical formulations as tablet binder, film-coating agent, etc.3 One possible structural element of HPC is depicted in Figure 1. * Corresponding author: (e-mail) [email protected]; (fax) +46-31-7763798. (1) Brandt, L. In Ullmann’s Encyclopedia of Industrial Chemistry; Campbell, F. T., Pfefferkorn, R., Rounsaville, J. F., Eds.; VCH Verlagsgesellschaft: Weinheim, 1986; p 461. (2) Do ¨nges, R. Br. Polym. J. 1990, 23, 315. 10.1021/ac0301604 CCC: $25.00 Published on Web 10/16/2003

© 2003 American Chemical Society

For optimal development and use of cellulose derivatives, their functional properties and the factors that determine these properties have to be as well characterized as possible. It is known that the properties of these modified polymers depend on several structural parameters such as the type of substituent, the degree of substitution, and the distribution of substituents.2,4,5 Consequently, there is a need for various independent analytical methods for characterization of the aforementioned parameters. The substituent content in commercial cellulose ethers can be determined by a number of methods. One method widely employed for nonionic derivatives is The United States Pharmacopeia (USP) method for HPC, where a chromic acid oxidation followed by a redox titration is utilized.6 An alternative for hydroxypropoxy content analysis is a modification of the Zeisel alkoxy reaction, which uses a hydroiodic acid treatment followed by a gas chromatographic (GC) method.7-10 Among the methods used for determination of the substituent content in cellulose derivatives, those based on nuclear magnetic resonance (NMR) spectroscopy are considered to be the more accurate ones.10-14 In addition, 1H as well as 13C NMR are applied for determination of the substituent distribution within the anhydroglucose unit (AGU), i.e., the relative substitution at O-2, O-3, and O-6, respectively.10,15-17 However, one restriction to the practical use of the NMR technique is the limited solubility of intact cellulose (3) Kibbe, A. H., Ed. Handbook of Pharmaceutical Excipients, 3rd ed.; American Pharmaceutical Association and Pharmaceutical Press: Washington, DC, 2000; p 665. (4) Alvarez-Lorenzo, C.; Castro, E.; Go´mez-Amoza, J. L.; Martinez-Pacheco, R.; Souto, C.; Concheiro, A. Pharm. Acta Helv. 1998, 73, 113. (5) Mischnick, P.; Heinrich, J.; Gohdes, M.; Wilke, O.; Rogmann, N. Macromol. Chem. Phys. 2000, 201, 1985. (6) The United States Pharmacopeia 25, The National Formulary 20; United States Pharmacopeial Convention Inc.: Rockville, MD, 2002; p 2564. (7) The Japanese Pharmacopoeia, XIV ed. Society of Japanese Pharmacopoeia: Tokyo, 2001; p 940. (8) Miller, D. L.; Samsel, J. G.; Cobler, J. G. Anal. Chem. 1961, 33, 977. (9) Hodges, K. L.; Kestes, W. E.; Widerrich, D. L.; Groves, J. A. Anal. Chem. 1979, 51, 2172. (10) Alvarez-Lorenzo, C.; Lorenzo-Ferreira, R. A.; Gomez-Amoza, J. L.; MartinezPacheco, R.; Souto, C.; Concheiro, A. J. Pharm. Biomed. Anal. 1999, 20, 373. (11) Ho, F. F.-L.; Kohler, R. R.; Ward, G. A. Anal. Chem. 1972, 44, 178. (12) Kimura, K.; Shigemura, T.; Kubo, M.; Maru, Y. Makromol. Chem. 1985, 186, 61. (13) Lee, D.-S.; Perlin, P. S. Carbohydr. Res. 1982, 106, 1. (14) Andersson, T.; Richardson, S.; Erickson, M. Pharmeuropa 2003, 15, 271. (15) Reuben, J. Carbohydr. Res. 1986, 157, 201. (16) Tezuka, Y.; Imai, K.; Oshima, M.; Chiba, T. Carbohydr. Res. 1990, 196, 1. (17) Kern, H.; Choi, S. W.; Wenz, G.; Heinrich, J.; Ehrhardt, L.; Mischnick, P.; Garidel, P.; Blume, A. Carbohydr. Res. 2000, 326, 67.

Analytical Chemistry, Vol. 75, No. 22, November 15, 2003 6077

Figure 1. Schematic picture of a possible structure of hydroxypropylated cellulose.

derivatives in common NMR solvents. This problem can be circumvented by subjecting the polymer to pretreatments such as hydrolysis,12,13,15,18 methanolysis,15,19 or acetylation,16,20 which allows most cellulose derivatives, no matter their degree of substitution, to become soluble in the NMR solvents. The distribution of substituents on a monomer level has also been successfully studied by GC in combination with mass spectrometry (MS)5,21,22 after sample preparations that involve derivatization and complete degradation of the polymer, e.g., standard methylation analysis23,24 or reductive cleavage.24,25 However, properties such as solubility, viscosity, biodegradability, or formation of aggregates are generally thought to correlate more strongly with the distribution of substituents along the cellulose chain. This parameter has been studied after degradation of the polymer, either chemically5,22,26 or enzymatically,27-30 followed by analysis of the hydrolysis products by techniques such as SEC, NMR, or matrixassisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). Whereas SEC and NMR are well established, MALDI-TOF MS is a relatively new technique employed in the area of cellulose derivatives and related compounds. There are a few works published on this issue; e.g., Mischnick and Kuhn studied the substitution pattern in methyl amylose after partial chemical degradation by MALDI-TOF MS.31 A similar approach has been applied on enzymic hydrolysates of methylated pectins32,33 and carboxymethyl cellulose.34 (18) Parfondry, A.; Perlin, A. S. Carbohydr. Res. 1977, 57, 39. (19) Reuben, J. Carbohydr. Res. 1987, 161, 23. (20) Tezuka, Y.; Imai, K.; Oshima, M.; Chiba, T. Macromolecules 1987, 20, 2413. (21) Lindberg, B.; Lindquist, U.; Stenberg, O. Carbohydr. Res. 1987, 170, 207. (22) Arisz, P. W.; Kauw, H. J. J.; Boon, J. J. Carbohydr. Res. 1995, 271, 1. (23) Bjo ¨rndal, H.; Hellerqvist, C. G.; Lindberg, B.; Svensson, S. Angew. Chem., Int. Ed. 1970, 9, 610. (24) Mischnick, P. Macromol. Symp. 1995, 99, 3. (25) Rolf, D.; Gray, G. R. J. Am. Chem. Soc. 1982, 104, 3539. (26) Gohdes, M.; Mischnick, P. Carbohydr. Res. 1998, 309, 109. (27) Wirick, M. G. J. Polym. Sci. 1968, 6, 1965. (28) Horner, S.; Puls, J.; Saake, B.; Klohr, E.-A.; Thielking, H. Carbohydr. Polym. 1999, 40, 1. (29) Saake, B.; Horner, S.; Kruse, T.; Puls, J.; Liebert, T.; Heinze, T. Macromol. Chem. Phys. 2000, 201, 1996. (30) Richardson, S.; Lundqvist, J.; Wittgren, B.; Tjerneld, F.; Gorton, L. Biomacromolecules 2002, 3, 1359. (31) Mischnick, P.; Kuhn, G. Carbohydr. Res. 1996, 290, 199. (32) Ko ¨rner, R.; Limberg, G.; Mikkelsen, J.; Roepstorff, P. J. Mass Spectrom. 1998, 33, 836. (33) Limberg, G.; Ko ¨rner, R.; Buchholt, H.; Christensen, T.; Roepstorff, P.; Mikkelsen, J. Carbohydr. Res. 2000, 327, 321.

6078

Analytical Chemistry, Vol. 75, No. 22, November 15, 2003

In this work, certain structural characteristics of the substitution in two HPC samples from different suppliers were investigated and compared by means of various analytical techniques. The objectives were to study the potential and limitations of each technique for analysis of the substitution of modified polysaccharides. The results obtained were evaluated with respect to variations in substitution between the two HPC samples, e.g., substituent content or amount of unsubstituted glucose, factors that could correlate with differences in functional properties of the polymers. To estimate the degree of substitution (D.S., i.e., average number of substituted hydroxyl groups in the monomer unit) and molar substitution (M.S., i.e., average number of moles of substitution groups per monomer unit), 1H and 13C NMR were employed, whereas MALDI-TOF MS was used to study the distribution of the number of hydroxypropyl groups per monomer unit. The amount of unsubstituted glucose units present in the polymers, which may reflect the heterogeneity of substitution, was determined by high-performance anion-exchange chromatography (HPAEC) with pulsed amperometric detection (PAD). The molar mass distribution of the two HPC samples, which is yet another important parameter that could correlate with variations in behavior, has been thoroughly studied by SEC-MALS in a previous work35 and was not investigated further here. EXPERIMENTAL SECTION Chemicals. Two HPC samples from different manufacturers were investigated, HPC LM from Nippon Soda Co. (Tokyo, Japan) and HPC LF from Hercules (Wilmington, DE). The M.S. values reported by the manufacturers were 3.3 for HPC LM and 3.5 for HPC LF. Girard’s T reagent (99%) and glucose were purchased from Sigma-Aldrich Chemie GmbH (Steinheim, Germany), trifluoroacetic acid (TFA) was from Fluka Chemie GmbH (Buchs, Germany), and pyridine and acetic anhydride were from Merck (Darmstadt, Germany). Procedures. For cloud point analysis, HPC samples (∼250 mg) were dissolved in 10 mM NaCl to a final sample concentration of 1.0% (w/w). The samples were slowly stirred for 4 days. Acetylation of HPC samples was carried out before NMR analysis according to Tezuka et al.16,20 A 150-mg sample of HPC (34) Karlsson, J.; Momcilovic, D.; Wittgren, B.; Schulein, M.; Tjerneld, F.; Brinkmalm, G. Biopolymers 2002, 63, 32. (35) Wittgren, B.; Porsch, B. Carbohydr. Polym. 2002, 49, 457.

was dissolved in an acetic anhydride (2.25 mL)/pyridine (0.75 mL) mixture. The mixture was allowed to reflux for 3 h before dialysis of the acetylated products against water in Spectra/Por dialysis membranes, molecular weight cutoff (MWCO) 10 000 (Spectrum Laboratories, Rancho Dominguez, CA). Finally, the products were freeze-dried overnight. Complete degradation of HPC to monomers was accomplished by hydrolysis of HPC (5 mg) in 2 M TFA (1 mL) at 120 °C for 16 h. The acid was removed by evaporation. Subsequently, the residue was redissolved in water and finally freeze-dried overnight. The hydrolysis reaction was followed at regular time intervals by MALDI-TOF MS and HPAEC-PAD analysis. For HPAEC-PAD experiments, the dried products were dissolved in 1 mL of water. For analysis by MALDI-TOF MS, the hydrolysis products were derivatized with Girard’s T reagent for introduction of a cationic site.36,37 A solution of Girard’s T reagent with a concentration of 2 g/L was prepared by dissolving the reagent in a methanol/acetic acid mixture (8.5/1.5, v/v). Reagent solution (1 mL) was added to dried hydrolysate of HPC (1 mg), followed by derivatization at 75 °C for 3 h. The solvents were removed by evaporation, and the residue was dissolved in ethanol to a concentration of 10 g/L for direct MALDI-TOF MS analysis. Instruments. For the cloud point analysis, a Mettler Toledo FP900 Thermo system was used, consisting of a FP90 central processor connected to a FP81C measuring cell. The temperature interval for analysis was 30-70 °C. The light source for illumination of the samples was a 24-V, 2-W lamp. The predetermined definition of cloud point was the point where the initial transmission of light through the sample suspension was changed by 4%. The data were handled by an in-house software, Cloudpoint version 1.0. The 13C NMR analysis was carried out on a Varian 300-MHz Mercury instrument operating at 75.431 MHz. The NMR instrument was equipped with a 5-mm QNP probe. The samples were dissolved in Me2SO-d6, and the concentration was 7-10% (w/w). The temperature used was 80 °C. The 13C NMR measurements were carried out by using a flip angle of 45° and a pulse repetition time of 30 s. The spectral width was 17 000 Hz, and 1900-2000 transients were obtained. The protons were decoupled during the acquisition using WALTZ-16 modulation. A line-broadening factor of 2.7 Hz was used. The 1H NMR measurements were carried out on a Varian 500-MHz Inova instrument operating at 499.545 MHz. The NMR instrument was equipped with a 5-mm-i.d. probe. The samples were dissolved in Me2SO-d6, and the concentration was ∼2% (w/w). Usually the 1H NMR spectra were recorded both at 25 and at 75-80 °C. The temperature used was 80 °C. The 1H NMR measurements were carried out by using a flip angle of 45° with an acquisition time of 5 s and a delay time of 5 s. The spectral width was at least between -2 and 10 ppm referring to the solvent peak of DMSO (2.49 ppm) and 16 transients were obtained. A zero filling of at least 2-fold in size was applied, and a line broadening factor of 0.3 Hz was used. The chemical shift values (36) Naven, T. J. P.; Harvey, D. J. Rapid Commun. Mass Spectrom. 1996, 10, 829. (37) Momcilovic, D.; Brinkmalm, G.; Wittgren, B.; Wahlund, K.-G. Improved MALDI-TOF-MS analysis of Partially Depolymerisaed Methyl Cellulose by Reducing End Derivatisation. In 50th ASMS Conference on Mass Spectrometry and Allied Topics, Orlando, FL, 2002.

were referenced to the solvent signal of Me2SO-d6 (2.49 ppm for 1H and 43.5 ppm for 13C). The spectra were phased correctly prior to the integration. The area of peak a in Figure 4 was integrated between 2.20 and 1.60 ppm and peak b was integrated between 1.50 and 0.70 ppm. MALDI-TOF MS analysis was performed on a PerSeptive Voyager-DE STR (Applied Biosystems, Framingham, MA) equipped with an N2 laser, time lag focusing, reflector, and a tandem coupled microchannel plate detector in reflector mode. All mass spectra were acquired in reflector mode, with an accelerating voltage of 20 kV. Other source voltages and time lag were set to obtain maximum resolution. The guide wire and the detector low-mass gating were not activated. The laser intensity was held slightly above threshold. The mass spectra were acquired in positive ion mode and accumulated from 200 laser shots, collected at four different positions within the sample spot. 2,5-Dihydroxybenzoic acid (DHB) was used as matrix. DHB was dissolved 10 g/L in H2O and the HPC analyte 1 g/L in H2O. The matrix and analyte solutions were then mixed 4:1, and 1 µL of the mixture was applied on a MALDI sample plate and allowed to dry at reduced pressure. An HPAEC-PAD chromatographic system from Dionex (Sunnyvale, CA), consisting of a GS50 gradient pump, a CarboPac PA-1 guard and analytical column, and an ED50 electrochemical detector, was used for glucose determination. The electrochemical detector, with a gold working electrode versus a Ag/AgCl(sat) reference electrode, was operated at the following waveform: E1 ) 0.10 V (td ) 0.10 s, t1 ) 0.20 s), E2 ) -2.00 V (t2 ) 0.21 s), E3 ) 0.60 V (t3 ) 0.23 s), and E4 ) -0.10 V (t4 ) 0.24 s). The injection volume was 20 µL, and elution of the components in the hydrolysates was carried out isocratically with 75 mM NaOH at a flow rate of 0.25 mL/min. RESULTS AND DISCUSSION Aqueous solutions of cellulose derivatives have the typical feature of a reversed solution behavior, i.e., reduced solubility and eventually phase separation at increased temperatures. This effect is strongly influenced by the chemical composition of the derivative. The type of substituent, the degree of substitution, and the substituent pattern determine the hydrophobicity of a cellulose derivative and thus its phase behavior. One rather straightforward way to achieve a crude estimation of differences in hydrophobicity is to study the onset of phase separation (clouding) at increased temperature. This has been done for the two different HPC:s (Figure 2). For HPC LF, the clouding occurred already slightly above 40 °C whereas HPC LM required higher temperatures for phase separation. The reason for HPC LM having a higher clouding temperature compared with HPC LF is not known but it seems likely, especially since the obtained weight average molar masses for the two samples were quite similar (116 000 g/mol for HPC LM and 125 000 g/mol for HPC LF),35 that it reflects differences in substitution between the two cellulose derivatives. To investigate possible correlations between chemical structure and the clouding temperatures, various structural parameters such as molar substitution, degree of substitution, distribution of substituents on a monomer level, and amount unsubstituted glucose were studied. NMR Analysis. NMR analysis was carried out on acetylated HPC samples; i.e., the unsubstituted hydroxyl groups in the AGUs Analytical Chemistry, Vol. 75, No. 22, November 15, 2003

6079

Table 1. M.S. Values of HPC LM and HPC LF Determined by 1H NMR, USP Method, and MALDI-TOF MS M.S. values

Figure 2. Cloud point curves of HPC LM and HPC LF.

Figure 3. Structure of acetylated HPC.

Figure 4. 1H NMR spectrum of the acetylated HPC LF sample in Me2SO-d6 at 25 °C.

had been capped by an acetyl group.16,20 This prederivatization step facilitates 1H and 13C NMR analysis, as the solubility of the polymer in the NMR solvent (deuterated DMSO) is markedly increased over a wide range of degrees of substitution. Another advantage is that the cellulose derivatives are allowed to retain their polymeric form, thereby avoiding the otherwise necessary and sometimes troublesome hydrolysis or methanolysis step to be able to dissolve the polymer. Additionally, degradation of the polymer may result in the formation of anomers or bicyclical acetals, leading to serious spectral complications.38 Furthermore, the acetyl carbonyl carbon signal is highly sensitive to its position in the AGU, which allows direct determination of distribution of the hydroxypropyl groups in the glucose unit. 1H NMR spectroscopy (25 °C) was employed for determination of the M.S. of the two HPC samples. Figure 3 shows the structure of acetylated HPC, and Figure 4 presents the 1H NMR spectrum of the acetylated HPC LM sample. The M.S. value was calculated from the peak areas of the well-resolved proton signals from the methyl group of the introduced acetyl group (signal a) and the methyl group of the hydroxypropoxy group (signal b). The M.S. of HPC LM was estimated to 3.7, whereas a higher value, 4.1, was obtained for HPC LF (see Table 1). These values determined by the 1H NMR method are in good agreement with the values reported by the two suppliers, who both utilize the USP method for determination of the hydroxypropoxy content.6 To verify that the acetylation reaction had been driven to completion, measure(38) Lee, D. S.; Perlin, A. S. Carbohydr. Res. 1984, 126, 101.

6080 Analytical Chemistry, Vol. 75, No. 22, November 15, 2003

sample

supplier

HPC LM HPC LF

3.3 3.5

1H

NMR 3.7 4.1

MALDI

MALDI+GT

4.7 5.1

4.1 4.3

ments with acetylation times other than that reported (3 h) were also carried out. However, acetylations for 1 and 6 h resulted in essentially the same M.S. values as did acetylation for 3 h.14 Thus, there is no risk of underacetylation, which would result in too high M.S. values. Consequently, the 1H NMR measurements seem to be reproducible and robust with respect to acetylation time. For evaluation of the 1H NMR method for determination of the M.S., other magnetic field strengths (300 and 600 MHz, respectively) and higher temperatures (75-80 °C) were used for the 1H NMR measurements. According to the results obtained, no significant changes in the results could be observed. Furthermore, 1H NMR measurements were performed on the two HPC samples when the polymer was dissolved directly in CDCl3, thereby avoiding the somewhat time-consuming acetylation step. The M.S. values obtained were similar to those determined after acetylation; however, interference from the water signal made the integration uncertain and the reproducibility of the measurements was low. Thus, acetylation of the samples before the NMR analysis is the preferred method. The 13C NMR analysis gave information on the total D.S. as well as the D.S. of the individual positions in the AGU, O-2, O-3, and O-6, respectively. These values were obtained by integration of the signals from the carbonyl carbon signals of the acetyl groups at each unsubstituted position in the glucose ring and the acetyl group of the (oligo-)propylene oxide substituent end group;16 see Figure 5. If an acetyl group is directly attached to either O-2, O-3, or O-6, the 13C chemical shift of the acetyl carbonyl carbon would be different as compared to if it is attached to a hydroxypropoxy group. The 13C chemical shift of acetyl carbonyl carbon is even different if it directly attached to O-2, O-3, or O-6. It is therefore possible to distinguish the total amount of acetyl carbonyl carbon attached to O-2, O-3, or O-6 as well as to a hydroxypropoxy group. However, no signals from any acetyl carbonyl carbon at unsubstituted positions could be detected in the 13C NMR spectra, in neither the HPC LM nor the HPC LF sample, which indicates that all hydroxyl groups at O-2, O-3, and O-6 are substituted with at least one hydroxypropoxy group; i.e., the total D.S. is 3 and the individual D.S. of each position is 1 (see Table 2). These results seem to be fairly reasonable considering the rather high M.S. of the two HPC samples. However, it is well established that the reactivity of the hydroxyl groups in the AGU under conditions of hydroxypropylation follows the order O-2 > O-6 . O-3.10,13,15 Thus, it could be expected to distinguish at least some slight differences in D.S. between the different positions. One reason for not observing such a trend in this particular case is most probably the high molar substitution of the two samples. Another reason might be the low sensitivity of the 13C NMR analysis. As will be presented below, other techniques (MALDI-TOF, HPAEC-PAD)

Figure 5.

13C

NMR spectrum of the acetylated HPC LM sample in Me2SO-d6 at 80 °C.

Table 2. D.S. and DP Values of HPC LM and HPC LF Determined by 1H and 13C NMRa

sample

D.S.2

D.S.3

D.S.6

D.S.total

DP

glucose liberated (w/w%)

HPC LM HPC LF

1 1

1 1

1 1

3 3

1.2 1.4

1.8 4.5

a Amount of unsubstituted glucose units in HPC LM and HPC LF determined by HPAEC-PAD.

are able to detect small amounts of unsubstituted hydroxyl groups in these two HPC samples. There is also a chain-extending reaction with propylene oxide to consider, i.e., the average degree of polymerization (DP) of the (oligo-)propylene oxide chains. This DP value can be calculated from the ratio of the M.S. and the total D.S. values.16 The results obtained here are presented in Table 2, where it can be seen that HPC LF showed the highest DP value of the (oligo-)propylene oxide side chain, which is an obvious consequence of the higher M.S. value of this sample. MALDI-TOF MS Analysis. HPC LM and HPC LF were subjected to complete acid hydrolysis in order to study the composition of different monomer units by MALDI-TOF MS. The resulting monomers were also derivatized at the reducing end by carboxymethyltrimethylammonium chloride hydrazine (Girard’s T reagent, GT), which introduces a cationic site for the subsequent analysis by MALDI-TOF MS.36,37 The advantage of this cationization is that a 10-100-fold increase in detection sensitivity can be obtained. This is especially of value in the analysis of carbohydrates that normally ionize by adduction of metal ions with comparatively low efficiency. Furthermore, the cationic site results in a more uniform ionization efficiency of unsubstituted as well as substituted mono- and oligomers. Figure 6 presents the mass spectrum of an acid hydrolysate of HPC LM derivatized by GT, where monomers substituted with one to eight hydroxypropoxy groups can be detected. In Figure 7a, the monomer distributions of underivatized and GT-derivatized HPC LM are illustrated. The distribution of the GT-derivatized monomers is shifted toward lower M.S., and the relative abundances of these lower substituted monomers are higher in comparison with the underivatized monomers. Furthermore, a

Figure 6. MALDI-TOF mass spectrum of monomers GT-derivatized, obtained from complete hydrolysis of HPC LM.

peak corresponding to glucose is detected in the GT-derivatized hydrolysate, although the signals in this lower mass region must be considered with caution due to possible interferences from matrix ions. After thorough examination of the spectra it was concluded that no such interference occurred in this particular case. The distributions of the different substituted monomers present in the acid hydrolysates of HPC LM and HPC LF are shown in Figure 7b. HPC LM has a distribution that is shifted toward lower substituted monomers compared to HPC LF. This finding is in agreement with a higher M.S. value of HPC LF. The substituent distribution of HPC LF is also wider than that of HPC LM. The main reason for this is, of course, because of the higher M.S. value of HPC LF. However, when the respective relative widths were calculated (i.e., the standard deviation divided by the M.S. value), it turned out that the HPC LF distribution was 2-3% wider. Judging from the distributions presented in Figure 7b, the main contribution to this appears to originate from the species containing 0-2 substituents. For these species, HPC LF has almost the same intensity as HPC LM, suggesting HPC LF to be more heterogeneous in the substituent distribution within the glucose unit. This indication should be considered with caution, although it supports the data obtained by HPAEC-PAD (see below), since MALDI is not a true quantitative method and since the peak intensities of the species with 0-2 substituents are relatively low. Analytical Chemistry, Vol. 75, No. 22, November 15, 2003

6081

Figure 7. (a) Substituent distribution of underivatized and GTderivatized monomers of HPC LM and (b) substituent distributions of GT-derivatized monomers in HPC LM and HPC LF determined by MALDI-TOF MS. The values are the individual peak areas normalized to the total peak area of the respective distributions.

The M.S. values estimated from the MALDI-TOF MS analysis, both of GT-derivatized and of underivatized HPC, are considerably higher compared to those obtained from the suppliers as well as the NMR analysis; see Table 1. This is most likely due to discrimination effects in MALDI. There are several types of discrimination that may occur: analyte may segregate in the sample preparation, giving inhomogeneous incorporation in the

matrix crystals; carbohydrates containing no substituents are generally more demanding to analyze than those having substituents, giving too high M.S. values; and finally, in this m/z region, low-mass species normally discriminated against, again giving too high M.S. values. To avoid the two latter types of discrimination, the samples were GT-derivatized as mentioned above. The introduction of a cationic moiety by derivatization was expected to give a more uniform ionization efficiency. However, although the derivatization seems to reduce discrimination, it appears not to be completely eliminated. HPAEC-PAD Analysis. One way to study the heterogeneity of substitution in chemically modified cellulose polymers is to determine the amount of unsubstituted glucose units. A heterogeneously modified cellulose sample consists of more unsubstituted glucose than a homogeneously substituted sample with the same degree of substitution. Thus, the amount of unsubstituted glucose present in a polymer reflects the uniformity of the substituent distribution. In this work, the amount of unsubstituted glucose in the two HPC samples was determined by HPAEC-PAD after complete acid hydrolysis of the polymers. This approach turned out to give highly reproducible and selective results in a short time. Within a few minutes, a chromatogram was acquired from which glucose easily could be identified and quantified by means of glucose standards (Figure 8). The peaks that elute before the glucose peak in the chromatogram in Figure 8 correspond in all probability to substituted monomers. As there are no standards available for such hydroxypropylated monomers, it is not possible to identify these peaks. Furthermore, this method allows an evaluation of the degree of degradation of the polymer, as di- and oligosaccharides, if present, are readily detected. However, no disaccharides or higher oligomers could be seen in the chromatograms, which is evidence of a complete hydrolysis of the polymer to monomeric units. This is of importance in order to obtain an accurate value of the unsubstituted glucose content.

Figure 8. HPAEC-PAD chromatogram of an acid hydrolysate of HPC LM. 6082

Analytical Chemistry, Vol. 75, No. 22, November 15, 2003

According to the results presented in Table 2, 1.8% of HPC LM and 4.5% of HPC LF were converted into glucose. As the HPAEC-PAD results indicate the opposite, with HPC LF containing a higher amount of unsubstituted glucose units despite its higher M.S. value, it is suggested that HPC LF has a more heterogeneous distribution of hydroxypropyl groups along the cellulose chain compared with HPC LM. This speculation is in accordance with the indications that were given from the MALDITOF MS results. Even if the difference between the two samples might seem rather small, ∼2 unsubstituted glucose units/100 units compared with ∼4-5 unsubstituted units/100 units, it may very well be large enough to cause variations in the polymer properties. Most likely, HPC LF, with not quite double the amount of unsubstituted glucose units, more easily forms larger “blocks” of unsubstituted units. These blocks can in turn associate with each other to form aggregates. The tendency to form aggregates undoubtedly affects the functional properties of HPC and, thus, can be one explanation of any differences in behavior between HPC LM and HPC LF. The results from the glucose determination by HPAEC-PAD are in accordance with those obtained from the MALDI-TOF MS analysis. However, these results are not in full agreement with the 13C NMR analysis of the D.S. values of the two HPC samples. The fact that no signals from unsubstituted AGUs were observed in the 13C NMR spectra is therefore regarded to be an effect of the relatively low sensitivity of that technique. Thus, the D.S. values estimated by 13C NMR analysis have to be considered as rough approximations. CONCLUSIONS The present work illustrates the use of and need for various analytical techniques for characterization of substitution in cellulose derivatives. The techniques that have been evaluated in this work were shown to give different types of information and

to serve as complements to each other, which is essential in order to get a picture as complete as possible of the substitution. 1H NMR spectroscopy was employed for determination of the hydroxypropoxy content in HPC. The method turned out to be reproducible and less laborious compared with other general methods for hydroxypropoxy content such as the USP method. 13C NMR has the potential to estimate both the total D.S. and the D.S. of the individual positions in the AGU. However, it was too insensitive to detect the small differences in D.S. found for the investigated HPC samples. MALDI-TOF MS analyses of hydrolyzed HPC samples gave information on the distribution of the different substituted monomers present in the polymer, an approach that additionally may reflect the distribution of substituents along the polymer chain. One obvious limitation of this technique is the discrimination effects that affect all data based on signal intensity, making it more laborious to obtain quantitative results. Furthermore, it was demonstrated that determination of the glucose content in the polymer by means of HPAEC-PAD analysis was a fast and simple strategy to obtain information on the heterogeneity of the substituent distribution. The two HPC samples from different suppliers were shown to differ from each other both with respect to substituent content and substitution heterogeneity, factors that may very well explain differences in clouding behavior and other functional properties of these polymers. ACKNOWLEDGMENT Mr. Dane Momcilovic and Dr. Magnus Erickson are acknowledged for valuable scientific discussions. Received for review April 22, 2003. Accepted September 4, 2003. AC0301604

Analytical Chemistry, Vol. 75, No. 22, November 15, 2003

6083