Algal Toxins and other Natural Products.

1 Supporting Information 2 3 4 Quantitative 1 H NMR with External Standards: Use in Preparation of Calibration Solutions for Algal Toxins and other Natural Products. 5 6 7 Ian W. Burton, Michael A. Quilliam and John A. Walter* 8 9 Institute for Marine Biosciences 10 National Research Council of Canada 11 1411 xford Street, Halifax N.S. Canada B3H 3Z1 12

12 Contents 13 14 Supplementary discussion: Note: Numbering of references is the same as for the main manuscript. 15 SD1 Proportionality of NMR signal intensities to concentrations. 16 17 18 19 SD2. RF pulses and relaxation effects. SD3. Temperature effect on spin level populations. SD4. Probe Q-damping and effect on pulse width. SD5. Data processing and integration of spectra. 20 SD6 Summary of Conclusions. 21 22 Tables: Table S1. Response (integrated intensity I x p 360 ) of caffeine methyl resonances to 23 24 25 26 single 90 o pulses, probe T, for varying transmitter frequencies covering the full 15 ppm spectral width. Table S2. Caffeine solutions (4 mm) in H 2 at varying NaCl concentration, Probe T: Effects of probe Q-damping on integrated resonance intensities I(H 2 ), I(caff) and p 90. 27 Table S3. Comparison of product of solvent resonance integral per mole and p 90, for 28 H 2 in dilute caffeine solutions, neat CHCl 3 and neat CH 3 H, probe T. 29 Table S4. Effects of mistuning and mismatching of probe T on product of integrated 30 31 peak intensity I(H2) and p360, for the H 2 resonance of a caffeine solution containing trace GdCl 3. 32 Table S5. Concentrations of dilute solutes in H 2 measured by NMR against the 33 34 integral of the H 2 solvent signal, compared with determinations using gravimetry or NMR using an external standard. 2

35 36 37 Table S6. Caffeine, theophylline, arginine and saxitoxin (STX) solutions in H 2 : product of integral I(H 2 ) of solvent resonance (single 90 o pulse) and p 90 for probe X, over a 3 week period. 38 Table S7. Linearity of response: Caffeine average resonance intensity (3 samples, 39 methyl resonances) vs gravimetric concentration (mm), probe X. 40 Table S8. Sucrose Concentrations determined by NMR with probe X against five 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 external caffeine standards prepared gravimetrically, compared with gravimetric concentrations of same sucrose solutions. Figures: Fig. S1. Structures of compounds used for concentration determinations by NMR, including the shellfish toxins domoic acid, gymnodimine, okadaic acid and some of the Paralytic Shellfish Poisoning (PSP) toxins. Fig. S2. Effects of increased probe damping on signal intensity and p 90 for probe T. Fig. S3. Linearity of response and comparison of integration of same datasets by two different operators. Fig. S4. Concentrations of sucrose solutions determined by interpolation of 1 H NMR integrals against a calibration curve of caffeine integrals vs gravimetric concentration, probe X. Fig. S5. Concentration of theophylline solutions determined by NMR with a 90 o pulse against 4 mm standard caffeine solutions, vs gravimetric concentration of theophylline, probe X. Fig. S6. Concentration of theophylline solutions determined by NMR with 47 o pulse against 4 mm standard caffeine solutions, vs gravimetric concentration of theophylline, probe X. 3

59 60 SD1 Proportionality of NMR signal intensities to concentrations. We will restrict discussion to pulse Fourier-transformed spectra integrated in the 61 62 63 64 65 66 67 68 69 70 frequency domain. The strength I A of the 1 H NMR signal (per proton in a given chemical environment) produced by an analyte A, compared to that (I S ) of a standard S recorded under identical conditions, is proportional to the ratio N A /N S, where N A and N S are the respective numbers of 1 H nuclei in A and S, assumed to have the same geometrical distribution within the sensitive area of the probe. Where protons are in more than one chemically-distinct environment, independent measurements of I may be obtainable from one spectrum. As N A and N S are directly proportional to the concentrations C A and C S respectively, I A = k A C A and I S = k S C S and the molar ratio C A /C S may be determined directly from the intensity ratio if k A = k S. The experimental tests and discussion cover the steps necessary to ensure k A = k S when standards and analytes are in separate tubes. 71 72 SD2. RF pulses and relaxation effects 73 74 75 76 77 78 79 80 81 Rapid repetition of a one-pulse acquisition sequence produces partial saturation of resonances. This can be avoided by use of a flip angle << 90 o, 18 but a 90 o pulse with increased delay between pulses lessens the dependence of signal strength on precise setting of the pulse length. As the objective is to avoid saturation effects, the ultimate S/N per unit time will be similar no matter which alternative is adopted. The total time between pulses is normally the sum of a very short delay of a few µsec to allow recovery from the pulse, a fixed acquisition time ca 2-4 s, and a relaxation delay of 0-20 s. With 90 o pulses virtually complete relaxation (to within 0.7% of the maximum) occurs after a time 5T 1, where T 1 is the spin-lattice relaxation time for a given nucleus. T 1 can also be 4

82 83 84 85 measured if necessary for assurance of complete relaxation. The 90 o pulses should be short enough to provide uniform excitation over the full spectral width SW (i.e., reciprocal of pulse length >> SW). T 1 s are normally sufficiently long that relaxation during the recovery delay is negligible. 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 SD3. Temperature effect on spin level populations When using external standards for NMR quantitation, temperatures must be the same for both analyte and standard samples as each is assumed to have the same equilibrium Boltzmann distribution of spin level populations. For two spin (I = ½) levels in a magnetic field B o at high absolute temperature T A, such that γhb o /4π << kt A, the equilibrium spin population difference is closely approximated by: p 1 p 2 = p 1 γhb/2πkt A = p A where γ is the nuclear gyromagnetic ratio, and k is Boltzmann s constant. For a small change in temperature from T A to T B, (where T << T A ) the proportional change in population difference is therefore: ( p A p B )/ p A = T/T B Thus at room temperature (20 o C = 293 o K), a 1 o K increase of temperature would cause a 1/293 (0.34 %) decrease in spin level population difference, and a corresponding decrease in the detected magnetization. 101 102 SD4. Probe Q-damping and effect on pulse width 103 104 The effects of sample RF susceptibility and conductivity on the probe response ( Q factor of the tuned circuit) may be profound if the sample tube diameter is large 5

105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 compared to the separation of the probe coils. The effects can be visualized with the probe tuning display (RF bridge and swept frequency source) usually provided with current commercial NMR spectrometers, or by an external sweep generator and bridge connected to the probe. For all quantitative work with a separate external standard, the probe must be tuned and matched to the preamplifier (usually 50 Ω) for each sample, at the observing frequency. As probe tuning usually varies during sample spinning, samples should be static. A highly-conducting sample such as a salt solution will broaden the matched probe response corresponding to a lowered (damped) Q. As the intensity of the NMR signal is proportional to Q, 27 damping is a major factor preventing the condition k A = k S from being satisfied. However, increased damping also increases the 90 o (and 360 o ) pulse lengths (p 90 and p 360 respectively), and these readily measurable quantities can be used as a check and a correction for damping when the same coil provides both RF irradiation (at constant power) and detection of the nuclear magnetization, as in the probes used for our experiments. With fully protonated solvents, the probe damping may also be measured and compensated by integrating the solvent resonances of analyte and standard following a single 90 o pulse without presaturation, as the 1 H concentration for each solvent is calculable for dilute solutions from the known solvent densities. When RF irradiation and NMR detection are performed with the same coil, and all amplifiers are tuned and matched at the resonant frequency so that their characteristics are reproducible between samples, an inverse relation between p 90 and signal intensity is consistent with the principle of reciprocity. This states 28 that the emf ξ induced in an arbitrary conductor (coil) by time-dependent motion of a magnetic moment vector m at 6

127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 some point in its vicinity, is related to the magnetic field vector b that would be produced at that point by unit current in the conductor: ξ = -( / t)(b.m). The total emf is the sum of contributions ξ j from nuclear moments across the whole sample. 28 During RF pulse irradiation, at an arbitrary position j, p 90 1/(B 1j ) XY, where (B 1j ) XY is the component at point j of the alternating RF field B 1 in the XY plane perpendicular to the static field B 0 defined along the Z axis. If B 1j makes an angle θ to Z (which is also the direction of the equilibrium magnetization): p 90j = k 1 /[ B 1j Sin θ] The measurement of p 90 (or p 360, which may be determined more precisely), quantifies the effect of the medium (damping) on the magnitude of the RF magnetic field produced by the coil at a particular point j. Media of higher conductivity will result in longer p 360. The signal (emf) in the same coil, produced by precessing nuclei at the same position j, may be measured in response to a pulse p X producing arbitrary (but known) flip angle, usually but not necessarily 90 o. Following the pulse p X, the nuclear magnetization component in the XY plane, (m j ) XY, precessing at constant frequency ω about the Z direction, induces an emf ξ j given by: ξ j = - ( / t)(b j.m j ) = k 2 ( / t)(b 1j.m j ) as b j is in the same direction as B 1j. The vector b j is not time dependent. The time derivative of m j is: i(ωt + φ) ( / t)m j = iω (m j ) XY e where φ is an arbitrary phase angle. Therefore ξ j = iω m jxy e i(ωt + φ) k 2 B 1j Sinθ 7

150 151 152 153 154 155 156 157 158 159 160 161 162 The integrated signal intensity I j recorded after amplification, detection and Fourier transformation is proportional to the amplitude of ξ j : I j = k 3 ω m jxy k 2 B 1j Sinθ Therefore the product p 90j I j = { k 1 /[ B 1j Sin θ]}{k 3 ω m jxy k 2 B 1j Sinθ} = k 1 k 2 k 3 ω m jxy i.e. p 90j I j has a constant proportionality to m jxy at an arbitrary point j in the sample, provided that the flip angle p X is constant and the same coil is used for excitation and detection. RF susceptibility and conductivity effects which influence the coupling of the precessing nuclear magnetization to the coil will have the same effect on the radiofrequency field (produced by the same coil at the same frequency) that causes the excitation of the same parcel of nuclei. When a measurement of p 90 = p 360 /4 is made, the null at the 360 o pulse is a sample average, as the criterion is zero integral of the dispersive residual signal. Some nuclear magnetization elements are rotated by more, and some by 163 less, than 360 o due to B 1 inhomogeneities. When comparing different samples the 164 165 166 167 168 169 170 171 172 distribution of B 1 inhomogeneities will be the same as the samples extend well beyond the sensitive region of the coil. Magnetic susceptibility-induced differences in B 1 distribution should be negligible in these slightly paramagnetic or diamagnetic solutions, so the p 360 measurement should cover the same geometrical distribution of nuclear magnetization in different samples. The fact that any change in Ι due to dielectic and conductivity effects in the sample is exactly the same as the proportional change in 1/p 90, provides a direct means for compensation of probe damping effects when comparing different samples, and removes one of the major obstacles to quantitative use of NMR with separate external standards. 8

173 174 175 176 177 178 179 180 Probe damping may be minimized by use of a detection coil that is large in comparison to the sample tube diameter, or by reducing the concentration of conductive samples. An alternative approach (the ERETIC method 22,23 ) is to provide a small exponentiallydamped RF signal, at a frequency within the detected range, to an untuned (e.g. 13 C) coil in the NMR probe, and calibrate the area of the resulting peak to the resonance areas for a standard sample. It is claimed that this peak area will be influenced by probe damping to the same extent as 1 H resonances; however, for quantitative comparisons it would still be necessary to calibrate changes in flip angle. 181 182 SD5. Data processing and integration of spectra 183 184 185 186 187 188 189 190 191 192 193 Factors affecting the accuracy and precision in data processing of quantitative NMR spectra have been discussed in general 3,7,9,26 and in the context of natural products. 18 We will restrict discussion to Fourier-transformed spectra integrated in the frequency domain. Free induction decays (FID s) must be processed with apodization functions that leave the first point of the FID unaltered and do not cause excessive line broadening, e.g. a decreasing exponential corresponding to line broadening less than a typical linewidth, or a half-gaussian function having value 1.0 at zero time. Alteration of the first data point by apodization will change the integrated areas of the tranformed spectrum rendering quantitation impossible. Following Fourier transformation, baseline correction may be necessary, for which it is preferable to use reproducible means such as subtraction of linear or simple polynomial 194 segments over small regions of the spectrum. 18 Integration should be performed 195 numerically taking care to include all the area in the wings of individual resonances. 9

196 197 198 199 200 201 202 203 204 205 206 207 208 209 Baselines should be adjusted so that integral displays are horizontal at the start and end of the integrated range, and should be performed at a constant scaling factor for all samples. When lines are Lorentzian, the integral range ± R must be large compared to the linewidth ν at ½ height, to incorporate all the area that is significant above noise: e.g. to incorporate 99% or more of the area requires R/ ν > 24. 7 In practice, this restriction contributes little to inaccuracy in integration of spectra from natural products as linewidths are usually comparable across the spectrum and are only approximately Lorentzian in shape owing to residual field inhomogeneities. The inclusion of 1 H 13 C satellite resonances within particular integration ranges has to be considered on a caseby-case basis, 31 as their overlap with 1 H 12 C resonances varies with field as well as from one compound to another. Published studies of integration reproducibility with 1 H spectra of a drug 31 and a pure natural product 18 showed integration errors of 0.5% to 2%, repeatability within one laboratory being better than reproducibility between different laboratories. 31 210 211 SD6 Summary of Conclusions. 212 213 214 215 216 217 218 Quantitative NMR using precision tubes and external standards enables the direct measurement of analyte concentrations without contaminating the sample with internal standards or compromising subsequent use of the solutions with other techniques, such as LC - MS. The method is also advantageous for use with hazardous materials as samples may be kept in sealed tubes throughout. Expensive deuterated solvents can be avoided, overlap of standard with analyte resonances is not a concern, impurities in the standard are physically separated from the analyte, and standards may be prepared with greater 10

219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 accuracy. The 1 H spectrum provides in addition a measure of the analyte purity and integrity. Analytes and standards in different solvents or in salt solutions may be compared by quantitative corrections of Q-damping effects via ratios of analyte/standard p 360 values, or by measurements of solvent resonance integrals following a single 90 o pulse. The factors affecting the accuracy and precision of quantitative NMR using external standards have been explored and their relative influence is summarized in Table 2. Provided that all samples are accurately tuned and matched at the observing frequency, and that sample temperatures are constant, and that adequate relaxation delays are used, the major contributions to error arise from the accuracy and purity of the gravimetrically prepared external standards, the purity of the sample (particularly freedom from protoncontaining impurities producing resonances that overlap with the analyte, polymerization of the analyte, and macromolecules), subjectivity in the integration of spectra, and intermolecular NEs if presaturation is used to suppress resonances of non-deuterated solvents. Integration subjectivity is highly sample-dependent. With pure analytes having sharp well-separated resonances (e.g. caffeine) different experienced operators processing and integrating the same dataset can agree to better than 0.3%. Results with natural products that have more complex spectra and which may contain residual protonated impurities are influenced by the choice of peaks to integrate and baseline correction, and the only option may be to average the results obtained by two or more operators on the same dataset. Spectral overlaps with proton-containing impurities may sometimes be detected by 2D TCSY spectra. Impurities that adsorb the analyte will lead to broadening or 11

242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 indetectability of resonances and must be avoided. These problems also occur to the same extent when internal standards are used and may be further compounded by overlap of analyte and standard resonances. Intermolecular NE s are not accurately predictable, although as illustrated by the agreement between NMR concentration measurements and other methods (Table 1, Fig. 2) they are likely to be negligible within error for with most 1 H resonances in natural products. A practical strategy is to disregard peaks for quantitation if their intensity is significantly enhanced relative to others, and to employ structural knowledge of the molecule to decide which H atoms are likely candidates for this effect. The potential problems peculiar to NMR with external standards contribute little to the overall accuracy and precision. With single-coil RF irradiation and detection, differences in damping between analyte and standard may be corrected from the ratio of 360 o pulse lengths. A series of measurements with two probes and a variety of solvents and solutes showed that the product of the molar response of the solvent resonance and the 360 o (or 90 o ) pulse was constant, for a given probe, to better than 1%, the standard deviation being typically less than 0.5%. This error also includes any variation in sample tube diameter, and is consistent with the tolerances claimed by the manufacturer. ther possible sources of variation included within this figure are temperature effects on level populations, drift of total amplifier gain, digitizer response and integration. Constancy of the total response over a period of months has been demonstrated so it is possible to use the technique with confidence using periodic calibrations against standards. 12

263 264 265 266 267 268 Although we have worked with samples in non-deuterated solvents owing to the ultimate uses of the reference materials, the methods are more readily utilized with deuterated solvents as presaturation is unnecessary and field-frequency lock can be used. We have also shown that dilute solutions (mm range) of analytes in otherwise pure non-deuterated solvents may be quantified with a spectrometer having highly linear amplification stages and accurately calibrated receiver attenuation, by recording the 269 solvent-suppressed analyte signal with multiple 90 o pulses at high gain, against the 270 271 272 273 274 275 solvent signal from the same solution recorded from a single 90 o pulse at low gain. This method is faster as it does not require external standards, exact tuning, precisely matched sample tubes, damping corrections, or multiple samples, and it could be used with probes having separate irradiation and detection coils. Necessary precautions are to ensure dryness of solvents, and the same signal path in the instrument for measurements of solvent and analyte signals. 276 13

276 277 278 279 280 281 Table S1. Response (integrated intensity I x p 360 ) of caffeine methyl resonances to single 90 o pulses, probe T, for varying transmitter frequencies covering the full 15 ppm spectral width. Sample was 41.312 mm caffeine in D 2 with added GdCl 3 for rapid relaxation (linewidth of caffeine resonances ca 4 Hz, delay between pulses > 60s). Effect of varying transmitter offset on integrated peak intensity I x p360 Sample: Caffeine 41.312 mm in D2 plus trace GdCl3 Me resonances (pk 2, pk 3, pk 4) Single 90 deg pulses, no presat, p360 = 23.3 us, delay between pulses 60s, temp 20C Transmitter I x p360 of pk 2 I x p360 of pk 3 I x p360 of pk 4 offset (Hz) 2309.1 0.985 0.978 0.959 Transmitter on residual H peak 1809.1 0.951 0.950 0.953 Transmitter on pk 2 1309.1 0.946 0.960 0.949 809.1 0.951 0.947 0.949 2809.1 0.953 0.948 0.941 3301.9 0.953 0.948 0.945 3809.1 0.955 0.946 0.959 2309.1 0.954 0.960 0.959 4309.1 0.957 0.954 0.955 Distance from edge of SW: 4809.1 0.953 0.954 0.959 pk 4 at 430Hz 5309.1 0.952 0.660 0.000 pk 2 at 264Hz, pk 3 at 32 Hz 309.1 0.952 0.947 0.958-190.9 0.944 0.959 0.943-690.9 0.950 0.946 0.948-1190.9 0.949 0.950 0.954 Av: 0.954 0.953 0.952 SD: 0.009 0.009 0.006 Av excl 2309.1: 0.951 0.951 0.952 SD: 0.003 0.005 0.006 Numbers in italics not used in averages. Response uniform to 1.0% SD across 15 ppm (7508 Hz) SW to within 264 Hz of edge. This includes both filtering and pulse power contributions to non-uniformity of response. 14

281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 Table S2. Caffeine solutions (4 mm) in H 2 at varying NaCl concentration, Probe T: Effects of probe Q-damping on integrated resonance intensities I(H 2 ), I(caff) and p 90. NaCl conc p(90) I(H2) I(H2) x p(90) Av I(caff)/H Av I(caff)/H Av I(H2) xp90 mm (µs) x p(90) per mole H 0.00 5.38 58.89 634.0 5.712 1.00 5.36 59.33 636.0 4.770 25.57 5.731 5.00 5.39 59.17 637.5 4.879 26.28 5.744 10.00 5.53 57.43 634.5 4.708 26.01 5.717 20.00 5.67 56.64 641.7 4.621 26.18 5.782 50.00 6.06 52.33 634.2 4.343 26.32 5.714 100.00 6.60 48.58 641.3 4.019 26.52 5.778 Av: 637.0 26.15 5.740 sd: 3.3 0.33 0.030 Note: concentration of H 2 in a 100 mm NaCl solution at 20 o C is 0.9983 that of pure H 2 (CRC Handbook of Chemistry and Physics, 60 th edition, 1979, p D-261). 15

301 302 303 304 305 Table S3. Comparison of product of solvent resonance integral per mole and p 90, for H 2 in dilute caffeine solutions, neat CHCl 3 and neat CH 3 H, probe T. These tests were run four months later than those in Table S1 and used a different set of RF filters, so absolute intensities are not strictly comparable with those in Table S1. 306 Neat solvent I(solv)/H Av solv x p90 Sample concentration single scan per Mole (M) p(90) Caff 1mM, NaCl 1mM T&M 55.49 5.19 59.16 5.533 T&M 55.49 5.22 59.00 5.544 T&M 55.49 5.21 59.03 5.542 Caff 1mM, NaCl 100mM T&M 55.49 6.53 47.36 5.583 CHCl3 Neat T&M 12.50 5.27 6.60 5.563 MeH Neat T&M 24.69 5.83 11.83 5.593 NS=1 T&M 24.69 5.83 11.83 5.595 T&M 24.69 5.83 11.80 5.589 Av: 5.568 sd: 0.025 Caff 1mM, NaCl 1mM Mismatch 55.49 6.49 52.62 6.154 Caff 1mM, NaCl 100mM Mismatch 55.49 9.64 43.59 7.582 T&M = probe tuned and matched 16

306 307 308 Table S4. Effects of mistuning and mismatching of probe T on product of integrated peak intensity I(H2) and p360, for the H 2 resonance of a caffeine solution containing trace GdCl 3. Effect of mistuning and mismatching on integrated peak intensity I x p360 Sample: Caffeine 40.974 mm in H2 plus trace GdCl3 Probe T Single 90 deg pulses, no presat, delay between pulses 60s, temp 20C I (H2) is integrated intensity of H2 resonance plus caffeine peaks (I x p360) / p360 (us) I (H2) I x p360 (I x p360) tuned & matched Tuned and matched: 23.3 116.6 2718 1.00 Tuned, not matched: 24.7 100.9 2492 0.92 (match capacitor at one extreme of range) Tuned, not matched: 23.6 126.0 2973 1.09 Match capacitor at other extreme of range) Matched, not tuned: 36.0 103.8 3736 1.37 (gross mistuning) Matched, not tuned: 23.2 121.7 2822 1.04 (transmitter frequency at 1/3 of tuning display depth) 17

309 310 311 312 Table S5. Caffeine, theophylline, arginine and saxitoxin (STX) solutions in H 2 : product of integral I(H 2 ) of solvent resonance (single 90 o pulse) and p 90 for probe X, over a 3 week period. Day when Compound Conc NaCl conc data I(H2) p360 p90 p90*i(h2) recorded Caffeine 4 mm 0 1 97.2 77.0 19.25 1871 Caffeine 4 mm 1 mm 1 96.8 77.4 19.25 1872 Caffeine 4 mm 5 mm 1 96.8 77.0 19.35 1864 Caffeine 4 mm 10 mm 1 96.7 77.0 19.25 1861 Caffeine 4 mm 20 mm 1 96.9 77.0 19.25 1865 Caffeine 4 mm 50 mm 1 96.6 77.4 19.35 1868 Caffeine 4 mm 100 mm 1 95.8 78.3 19.58 1875 Theophylline 2 mm 0 2 95.5 78.2 19.55 1867 Theophylline 2 mm 0 2 95.4 77.9 19.47 1857 Theophylline 2 mm 0 2 95.4 77.9 19.48 1858 Theophylline 4 mm 0 2 95.4 77.9 19.48 1858 Theophylline 4 mm 0 2 95.5 77.8 19.45 1857 Theophylline 4 mm 0 2 95.3 77.9 19.47 1856 Theophylline 1mM 0 2 95.4 77.9 19.48 1858 Theophylline 1mM 0 2 96.3 77.9 19.48 1875 Theophylline 1mM 0 3 96.5 77.9 19.48 1879 USP Caffeine 4 mm 0 3 96.0 77.8 19.45 1868 USP Caffeine 4 mm 0 3 95.8 77.8 19.45 1864 USP Caffeine 4 mm 0 3 95.7 77.9 19.47 1862 sd Av all Caffeine, Theophylline: 1865 7.0 Av of solutions without NaCl: 1864 7.6 STX 0 6 94.5 77.9 19.47 1839 STX 0 6 94.5 77.9 19.47 1840 STX 0 6 95.3 77.4 19.35 1845 Arginine 4 mm 0 6 95.8 78.3 19.58 1875 Arginine 4 mm 0 6 96.0 78.2 19.55 1876 Arginine 4 mm 0 8 95.7 78.2 19.55 1871 STX 0 21 95.3 78.9 19.73 1879 STX 0 21 94.1 79.4 19.85 1867 STX 0 21 94.0 78.9 19.72 1854 Av all STX, Arginine: 1861 16.4 Av all solutions: 1864 10.8 18

312 313 Table S6. Linearity of response: Caffeine average resonance intensity (3 samples, methyl resonances) vs gravimetric concentration (mm), probe X. Caffeine: Signal intensity vs Gravimetric concentration, Probe X Grav conc Av integral verall Av SD (mm) ea. Sample 49.12 21085 21062 21089 30 21121 19.73 8398 8312 8366 47 8387 9.83 4150 4228 4199 43 4220 3.96 1676 1679 1674 6 1668 1.97 848 843 846 3 846 314 315 19

315 316 317 318 319 320 321 322 Table S7. Sucrose Concentrations determined by NMR with probe X against five external caffeine standards prepared gravimetrically, compared with gravimetric concentrations of same sucrose solutions. Peaks showing anomalous intensity attributable to presaturation (caffeine CH, two sucrose ranges) have not been used. Probe damping did not alter between solutions, so corrections for this were not needed. Caffeine std conc (mm) 1.97 3.96 9.83 19.73 49.12 Suc conc Gravimetric by caffeine sucrose conc Sucrose conc (mm) by NMR against each caffeine std Av suc conc calib. Curve (mm) Caff conc x Av I(suc) / Av I(caff) by NMR (mm) sd (mm) 48.72 49.47 48.96 49.33 48.71 49.04 0.35 48.78 48.73 19.95 20.26 20.05 20.20 19.95 20.08 0.14 20.02 19.74 9.649 9.797 9.697 9.769 9.648 9.71 0.07 9.72 9.68 3.819 3.878 3.838 3.867 3.819 3.84 0.03 3.89 3.94 1.799 1.827 1.808 1.822 1.799 1.81 0.01 1.87 1.93 20

322 323 324 325 326 327 328 329 330 331 332 Table S8. Concentrations of dilute solutes in H 2 measured by NMR against the integral of the H 2 solvent signal, compared with determinations using gravimetry or NMR using an external standard: S8(a) 40 mm caffeine; S8(b) domoic acid; S8(c) decarbomyl neosaxitoxin (dcne). For S8 (a) and S8(b) the integrals of the analyte and H 2 signals were taken from the same spectrum measured with probe T. The calculations assume negligible change in H 2 concentration (55.398 M at 20 o C) due to presence of analyte. Comparisons of domoic acid with caffeine standards were corrected for differences in p 360. Table S8(a) and S8(b): Sample Gravimetric Conc from SD Conc from SD conc (mm) H2 signal ext caffeine (mm) (mm) (a) Caffeine 40.974 40.80 41.14 41.77 Av: 41.24 0.49 (b) Domoic acid 10.99 11.14 10.90 10.77 10.84 10.90 10.91 10.68 10.85 10.76 10.85 10.68 10.93 Av: 10.82 0.17 10.88 0.04 Error incl. rel. error of caffeine stds: 0.28 333 334 335 336 337 338 For S8 (c), the integrals for the analyte were measured with a presaturation sequence at a high receiver gain setting (RG = 32) while the H 2 integral was measured with a single pulse sequence at low gain (RG = 1), using Probe X. The gain ratio was calibrated by comparing the integrals of a residual HD signal in D 2 (sealed sample) seven times at each RG setting, using a single pulse sequence with a 60 s delay. 21

339 Table S8(c). Decarbamoyl neosaxitoxin, probe X. Concentration (mm) of two decarbamoyl neosaxitosin (dcne) solutions determined against caffeine, with p90 correction, Probe X Range # Norm I*p90/scan sd Conc sd of means dcne A 1 0.670 0.016 2.25 2 0.636 0.026 2.14 3 0.639 0.034 2.15 Av 2.18 0.06 dcne B 1 0.659 0.033 2.21 2 0.646 0.037 2.17 3 0.660 0.005 2.22 Av 2.20 0.03 340 341 Av of dcne A and dcne B 2.19 0.04 Calibration of gain ratio for attenuator settings RG 32 / RG 1 The following attenuator ratios obtained with same sample of HD in D2, p360=84.0us, p90=21.0us Attenuator ratios: RG32/RG1 RG = 1 RG = 32 RG32/RG1 Dataset # Integ tox0323y 1,17 1.017 33.975 33.41 18,17 1.0058 33.975 33.78 tox0326y 1,3 1.0414 34.36 32.99 4,6 1.0374 34.403 33.16 7,8 1.0357 34.507 33.32 37,38 3.7186 121.89 32.78 39,40 3.6922 124.51 33.72 342 Av: 33.31 sd 0.37 Concentration (mm) of the same dcne solutions determined against solvent H2 No p90 correction needed as H2 and dcne in same sample Probe X Range # Norm I/scan sd Conc sd dcne A 1 0.0261 0.0006 2.30 0.05 2 0.0248 0.0010 2.19 0.09 3 0.0249 0.0013 2.20 0.12 Conc of dcne A: Av 2.23 0.06 sd of means dcne B 1 0.0257 0.0013 2.27 0.11 2 0.0252 0.0015 2.22 0.13 3 0.0257 0.0002 2.27 0.02 Conc of dcne B: Av 2.25 0.03 sd of means Av of dcne A and dcne B: 2.24 0.05 22

343 H H 3 C N CH 3 N N N H H N N N N CH 3 H H H H H H H CH 3 CH 3 Caffeine Theophylline Sucrose H H CH 3 H 3 C CH CH H H N CH N Domoic Acid H Gymnodimine H H 3 C H kadaic Acid CH 3 H CH 3 H H H H 3 C H CH 3 H R 1 R 2 R 3 R 4 344 345 346 347 R 1 R 4 H NH N NH 2 + STX H H H NE H H H GTX2 H H S - 3 NH 2 GTX3 H S - + H 2 N N NH 3 H H GTX1 H H S3- H GTX4 H S - 3 H R 2 R 3 GTX5 H H H dcstx H H H dcgtx2 H H S - 3 H dcgtx3 H S - 3 H NHS 3 Fig. S1. Structures of compounds used for concentration determinations by NMR, including the shellfish toxins domoic acid, gymnodimine, okadaic acid and some of the Paralytic Shellfish Poisoning (PSP) toxins. 348 23

348 H2 Integral/H TBI Av Caff Int/H (2-4) TBI 349 61 5.0 350 351 352 H2 Integral 59 57 55 53 51 49 y = -0.1100x + 58.97 R 2 = 0.973 Integral 4.8 4.6 4.4 4.2 4.0 y = -0.00786x + 4.709 R 2 = 0.973 353 354 47 0 20 40 60 80 100 120 [NaCl] mm 3.8 0 20 40 60 80 100 120 [NaCl] mm 355 356 a. y = -(0.1100 ± 0.0082)x + 58.97 ± 0.35 b. y = -(0.00786 ± 0.00065)x + 4.709 ± 0.030 R 2 = 0.973 R 2 = 0.973 357 358 359 360 361 362 363 90 deg (us) 6.8 6.6 6.4 6.2 6.0 5.8 5.6 5.4 5.2 TBI 90 deg pulse length (us) y = 0.01256x + 5.378 R 2 = 0.9925 5.0 0 20 40 60 80 100 120 [NaCl] mm 364 365 366 367 368 369 370 c. y = (0.01256 ± 0.00050)x + 5.378 ± 0.0214 R 2 = 0.9925 Fig. S2. Effects of increased probe damping on signal intensity and p 90 for probe T: Concentration dependence of: a. I(H 2 ); b. I(caff); c. p 90 for 4 mm caffeine solutions containing increasing concentrations of NaCl. 24

370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 Integral Integral 25000 20000 15000 10000 5000 0 25000 20000 15000 10000 5000 y = (430.4 ± 1.53)x 25.8 ± 36.9, R 2 = 0.99996 0 Caffeine integral vs gravimetric concentration: perator B y = 430.4x - 25.8 R 2 = 0.99996 0 10 20 30 40 50 60 [Caffeine] mm Caffeine integral vs gravimetric concentration: perator A y = 429.5x - 32.7 R 2 = 0.99997 0 10 20 30 40 50 60 [Caffeine] mm y = (429.5 ± 1.28)x 32.7 ± 31.1, R 2 = 0.99997 Fig. S3. Linearity of response and comparison of integration of same datasets by two different operators: Integral of caffeine methyl resonances (in H 2 solution) vs gravimetric concentration (av. of 3 solutions at each concentration, arbitrary integral units, integrated on same scale, probe X). The preamplifier used in these experiments was subsequently replaced. 25

401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 Caffeine integral a. y = (429.5 ± 1.28)x 32.7 ± 31.1, R 2 = 0.99997 NMR conc. of sucrose (mm) 25000 20000 15000 10000 5000 60 50 40 30 20 10 Caffeine integral vs gravimetric concentration: calibration curve 0 y = 429.5x - 32.7 R 2 = 0.99997 0 10 20 30 40 50 60 Caffeine gravimetric conc (mm) Sucrose quantitation by caffeine calibration curve y = 0.9976x - 0.0126 R 2 = 0.99996 0 0 10 20 30 40 50 60 Gravimetric conc. of sucrose (mm) b. y = (0.9976 ± 0.0038)x 0.0126 ± 0.0913, R 2 = 0.99996 Fig. S4. Concentrations of sucrose solutions determined by interpolation of 1 H NMR integrals against a calibration curve of caffeine integrals vs gravimetric concentration, probe X: a. Caffeine calibration curve. b. Sucrose concentrations by NMR from interpolation of caffeine calibration curve, vs gravimetric sucrose concentration. See Table S7. 26

440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 Theo NMR conc against caff (mm) a. y = (1.0012 ± 0.0040)x, R 2 = 0.99976 Theo NMR conc against caff (mm) Theophylline NMR conc vs gravimetric conc 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Theophylline NMR conc vs gravimetric conc 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 y = 1.0012x R 2 = 0.99976 0 1 2 3 4 5 Gravimetric Conc (mm) y = 0.9926x + 0.0261 R 2 = 0.99986 0 1 2 3 4 5 Gravimetric Conc (mm) b. y = (0.9926 ± 0.0044)x + 0.0261 ± 0.0017, R 2 = 0.99986 Fig. S5. Concentration of theophylline solutions determined by NMR with a 90 o pulse against 4 mm standard caffeine solutions, vs gravimetric concentration of theophylline, probe X: a. Linear regression through origin and data points. b. Linear regression through data points only. 27

471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 Theo NMR conc against caff (mm) a. y = (0.9969 ± 0.0072)x, R 2 = 0.9992 Theo NMR conc against caff (mm) 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Theophylline NMR conc vs gravimetric conc determined against 4 mm caffeine, 47 deg pulse y = 0.9969x R 2 = 0.9992 0.0 1.0 2.0 3.0 4.0 5.0 Gravimetric Conc'n (mm) Theophylline NMR conc vs gravimetric conc determined against 4 mm caffeine, 47 deg pulse y = 1.0034x - 0.0170 R 2 = 0.9993 0.0 1.0 2.0 3.0 4.0 5.0 Gravimetric Conc'n (mm) b. y = (1.0034 ± 0.0075)x 0.0170 ± 0.0173, R 2 = 0.9993 Fig. S6. Concentration of theophylline solutions determined by NMR with 47 o pulse against 4 mm standard caffeine solutions, vs gravimetric concentration of theophylline, probe X: a. Linear regression through origin and data points. b. Linear regression through data points only. 28