TECHNICAL NOTE

Impact of the deuterium isotope effect on the accuracyof 13C NMR measurements of site-specific isotope ratiosat natural abundance in glucose

Alexis Gilbert & Virginie Silvestre & Richard J. Robins &

Gérald S. Remaud

Received: 18 June 2010 /Revised: 29 July 2010 /Accepted: 24 August 2010 /Published online: 10 September 2010# Springer-Verlag 2010

Abstract The application of isotope ratio methods inauthenticity and traceability relies on the accuracy androbustness of the methodology employed. An unexpectedsource of error has now been identified, which canintroduce major and variable inaccuracies into the determi-nation of site-specific isotope ratio measurement byquantitative 13C NMR spectrometry if not correctlycontrolled. This is the isotope chemical shift effect, whichcomes into play when hydrogen atoms in the targetmolecule enter into exchange with deuterated water presentat trace levels in the deuterated solvent used as thefrequency lock. Even at a level of contamination as lowas 0.02%, an error of 5‰ can be introduced, fivefold therequired accuracy of 1‰. How to avoid this source of erroris discussed.

Keywords Isotopic 13C NMR . 13C isotope ratio accuracy .

Site-specific 13C isotope ratio . Isotope chemical shifteffects

Introduction

Isotope ratio determination in organic molecules is widelyapplied within several areas, notably food authentication,traceability, environmental studies, physiology and metabo-lism. The application of authentication and/or counterfeit

detection has a profound impact in the food, pharmaceuticaland cosmetics industries, in which counterfeiting anddeliberate mislabelling are serious and widespread problems.They even occur in the illegal manufacture of drugs ofabuse [1]. Verifying the authenticity of merchandise fromraw materials through to the finished products is thus ofmajor concern for both consumers and manufacturers, andseveral approaches have been developed in order to validatethe genuineness of a given product or process.

Among the analytical tools available, isotope ratiomeasurements are especially valuable, since the atomiccomposition of the target molecule(s) per se is probed [2].The isotope signature can be obtained by isotope ratio massspectrometry (irm-MS), but this technique provides only amean value for all the atoms of a given element present. Amore complete analysis, however, can be obtained usingquantitative isotope NMR spectroscopy, as this enables thedetermination of site-specific isotope ratios at naturalabundance. Thus, all resolved atoms in the molecule areindependently assessed. The exploitation of 2H NMR tocharacterise the origin of molecules has become widelyused since 1981 when Martin et al. proposed the SNIF-NMR method (site-specific natural isotopic fractionationstudied by nuclear magnetic resonance) for ethanol [3].

The logical extension, isotopic 13C NMR to measure thesite-specific 13C/12C isotope ratios has, however, provedchallenging due to the high level of trueness and precision(a clear definition of these terms is described in Ref. [4])required: of the order of 1‰ on the δ scale. In other words,the methodology should detect a difference in the signalareas of only 0.1% in a repeatable manner [5]. This isespecially critical when absolute values need to be determinedon either the same or different samples analysed severalmonths apart. The predominant source of inconsistency and

A. Gilbert :V. Silvestre : R. J. Robins :G. S. Remaud (*)Elucidation of Biosynthesis by Isotopic Spectrometry Group,Unit for Interdisciplinary Chemistry: Synthesis, Analysis,Modelling (CEISAM), CNRS–University of Nantes UMR6230,2 rue de la Houssinière, BP 92208,Nantes 44322 Cedex 3, Francee-mail: gerald.remaud@univ-nantes.fr

Anal Bioanal Chem (2010) 398:1979–1984DOI 10.1007/s00216-010-4167-9

inaccuracy was recognised as the homogeneity of the 1Hdecoupling over the whole range of proton frequency [6].Once this major criterion was satisfactorily resolved, othersignificant sources of inaccuracy were tackled: line shape,resolution of the satellite peaks, purity and/or impurityprofile of the samples under study [7]. The development ofthe method has now sufficiently progressed that it may beused routinely, and applications to such problems as thecounterfeiting of pharmaceuticals [7], the detection of fraudin pineapple juices and tequila [8] and the origins of ethanol[9], glycerol [10] and vanillin [11] are well advanced.However, as the technique moves from the research to theanalytical laboratory, it is increasingly crucial that potentialsources of error are minimised and that the method is robustand well characterised.

We have recently identified an additional unexpectedand potentially variable source of error: impurities in thesolvent used for spectral acquisition. Thus, traces ofdeuterated impurities in the solvent may carry out a slowexchange with hydrogen positions in the analyte. This leadsto a level of 13C–2H interaction that is not present in theoriginal analyte and the associated chemical shift isotopeeffect. This NMR phenomenon is well described in 13Cspectra for the hydroxyl chemical function [12, 13], and,generally, the resonance frequency of the 13C signal fromthe 13C–2H is more shielded. Both 13C signals of thespecies C–O2H and C–OH are observable, with intensitiesreflecting the respective population of each species. Theseextra signals, which are consistent with the isotopic effecton resonance frequency and in practice appear to the rightside of the main signal, are visible in the NMR spectrumonly if a slow exchange (on the NMR time scale) occurswithin the hydroxyl group between the OH and O2H forms.The isotope substitution effect may be detected on thecarbon directly linked to the hydroxyl group and also onremote carbons, described as the β and γ isotope chemicalshift effects, respectively. The associated chemical shiftchange Δ13C is defined as (δ13COH–δ

13CO2H), where δ13C

is the carbon chemical shift (in ppm) in the moleculecontaining OH and O2H, respectively. Δ13C is larger forthe β (two bonds) than the γ (three bonds) effect: Δ2C(2H)(or Δβ) and Δ3C(2H) (or Δγ), respectively [14].

Since the method of isotopic 13C NMR spectrometry relieson the accurate determination of the areas under the peaks inthe spectrum, this phenomenon could have a radicalinfluence on the quality of the isotopic 13C NMR spectrum.In the present communication, we demonstrate that thisphenomenon does indeed have a substantial impact. First, wereport the actual observations made when studying a glucosederivative that contains free OH groups. Secondly, weevaluate the outcome of this effect on the accuracy of themeasurements made. Finally, we suggest how to avoid thissource of error.

Experimental

Chemicals and materials

1,2-O-Isopropylidene)-α-D-glucofuranose (MAGF) usedfor isotopic 13C NMR measurement (preparation (a), seebelow) was synthesised as described previously [15].MAGF used for preparation (b) (see below) was purchasedfrom Fluka (www.sigma-aldrich.com). Deuterium oxide(2H2O), hexadeuterated dimethyl sulphoxide (DMSO-d6),which contains no residual deuterated species, and hexadeu-terated acetone (acetone-d6), which contains 0.02% 2H2O+H2HO, were purchased from Eurisotop (www.eurisotop.fr).The non-deuterated solvents were purchased from VWR(www.vwr.com).

NMR analysis

Samples were prepared according to one of three protocols.Preparation (a): MAGF (200 mg) was dissolved in DMSO(300 μL) and acetone-d6 (300 μL) was added. Preparation(b): MAGF (50 mg) was dissolved in DMSO-d6 (700 μL)and 2H2O (10 μL) was added. Preparation (c): 150 mgMAGF was dissolved in 300 μL DMSO-d6 and 400 μLmethanol. In all cases, the solution obtained was then filteredinto a 5 mm o.d. NMR tube (535 pp, Wilmad). Preparation(b) corresponds to a more diluted MAGF solution than inpreparation (a), in order to avoid the aggregation effect thatoccurs in a highly concentrated solution. Thus, what ismeasured is solely due to the isotope chemical shift effect(see “Results and discussion”).

13C NMR spectra were recorded using a Bruker DRX500-MHz spectrometer (preparation (a)) or a DPX 400-MHzspectrometer (preparation (b)) fitted with a 5 mm i.d. 13C/1Hdual probe carefully tuned at the recording frequency of125.76 or 100.64 MHz, respectively. The temperature of theprobe was set at 303 K for preparations (a) and (b) and at308 K for preparation (c), and spectra were acquired usingthe conditions described previously [15].

Results and discussion

Isotope shift in 13C NMR natural abundance spectraof glucose derivative MAGF

Glucose is a major raw material for the industrial productionof a wide range of products by fermentation. It is the primarycarbon reserve (in the form of starch and/or sucrose) in manyplants, and its 13C distribution, which is non-statistical,reflects the type of metabolism employed by the plant in itsbiosynthesis [15, 16]. It has previously been shown thatisotope fractionation from glucose to other metabolites

1980 A. Gilbert et al.

provides information crucial to understanding how isotoperedistribution occurs during fermentation [17, 18].

However, it is not possible to study the site-specific 13Cdistribution directly on glucose by 13C NMR because of theoccurrence of the α and β isoforms due to mutarotationalequilibrium. This difficulty has been overcome by designingan appropriate derivative of glucose, 3,5,6-triacetyl-1,2-O-isopropylidene-α-D-glucofuranose (TAMAGF) [15]. Initially,

MAGF, which as can be seen from Fig. 1 requires only twosteps of synthesis, was seen as a good candidate. However,as shown in Fig. 2, small “parasite” peaks appeared in the13C NMR spectrum when MAGF was dissolved in an equalmixture of DMSO and acetone-d6 (preparation (a), see“Experimental”). At first, these were thought to be due toan impurity, but a thorough investigation of the product(TLC, HPLC, GC, 1H NMR, and 13C NMR) failed to

5

4

3 2

1

O

OH

OH

OH

OH

6HO

345

O2

1 OH

OH

OH

6HO OH

345

O2

1 OH

OH

OH

6HO

O

5

4

3 2

1

O

OH

OH

OH

6HO

6

5

4 3

2

OO

O

1OHO

O

6

5

4 3

2

OO

O

1OAcO

O

2

2 2

D D

Fig. 1 Scheme of the protocolused for the derivatisation ofglucose in order to obtain δi

13Con each site of the glucosemolecule using TAMAGF asmolecular probe (modified fromRef. [15]). As can be seen, in themolecular structure of MAGF,1,2-O-Isopropylidene)-α-D-glucofuranose, there are threefree hydroxyl groups that canpotentially exchange hydrogenwith the deuterium of thesolvent. Numbering is that ofglucose conventional notation

Deuterium isotope effect on the accuracy of 13C NMR measurements 1981

identify any “new” impurity. Subsequently, the observationthat these extra peaks were only visible beside the signals ofthe OH-bearing carbons led us to consider that theseparasites were related to the isotopic chemical shift effect.

To confirm this hypothesis and characterise the origin ofthe phenomenon, we deliberately added 2H2O to a newpreparation of MAGF: DMSO-d6+

2H2O (preparation (b),see “Experimental”), conditions in which the full isotopechemical shift effect is observed (Fig. 3). For carbons 5 and6 (see Fig. 1 for numbering), four peaks are visible due tothe β and γ effects. These positions bear an OH group(β effect) and neighbour another OH in the α position(γ effect). Following the same rules, carbon 4 is subjectedto a double long-range γ effect, while carbons 2 and 3 areeach subjected to only one effect, γ and β, respectively.Carbon 1 is free of these effects since it has no adjacent OHgroup. This latter observation strongly supports the conclu-sion that the “extra peaks”, as shown in Fig. 2, are due to

the β isotope chemical shift effect. The signalcorresponding to the γ effect is not detectable because ofthe proximity of the main signal. Table 1 summarises thevalues of Δ13C for β and γ effects for both preparations (a)and (b): similar values are found, confirming that the effectobserved in preparation (a) is due to the presence of thechemical shift isotope effects.

Impact of the chemical shift isotope effect on the accuracyof the site-specific 13C isotope ratio measurements

As outlined in the “Introduction”, to observe such anoccurrence requires that two phenomena take place simulta-neously: (1) the presence of “free 2H atoms” and (2) a slowexchange of 1H and 2H atoms within the OH groups. Twoprocesses can be put forward to explain the latter. First,DMSO is a well-known solvent for slowing down chemicalexchanges [19], an effect associated with high viscosity and

74.0 73.8 73.6 69.2 69.0 68.8 64.2 64.0 63.8 63.673.4 68.6 68.4 63.4

C-3 C-5 C-6

Fig. 2 Zoom on the 13C NMR (1H decoupled) signals of carbon atomslinked to hydroxyl groups in MAGF (preparation (a): 200 mg in aDMSO/acetone-d6 (300 μL/300 μL) mixture, see “Experimental”).The rings delineate the presence of the “extra” peaks, due to the

chemical shift isotope effect. The other small peaks surrounding themain signal are the 13C satellites. Numbering is that of glucoseconventional notation (see Fig. 1)

C-1 C-2 C-4

C-3 C-5 C-6

105.4 105.3 105.2 85.6 85.5 85.4 81.0 80.9 80.8 80.7

74.1 74.0 69.3 69.2 69.1 64.464.6 64.564.3

β

γ

Fig. 3 13C-NMR (1H decoupled)signals of each carbon of MAGF(preparation (b): 50 mg in aDMSO-d6/

2H2O (0.7 mL/10 μL),mixture, see “Experimental”).Carbon atoms 3, 5 and 6 showsecondary chemical shift isotopeeffects (β effect). Carbon atoms 2,3, 4, 5 and 6 show tertiarychemical shift isotope effects(γ effect). These are indicated forcarbon no. 6 only. Numbering isthat of glucose conventionalnotation (see Fig. 1)

1982 A. Gilbert et al.

concentration: conditions met in our work. Secondly, MAGFcould form aggregates from a concerted self-assembly, asfound for similar compounds in gel forms [20].

The former point, however, is more intriguing.Where do thefree 2H atoms come from? The only possibility is thedeuterated solvent used for the field/frequency stabilisation(“lock”) of the NMR spectrometer. According to theinformation given by the manufacturer/supplier of deuteratedsolvents for NMR purposes, products may contain a mixtureof 2H2O and H2HO, with 0.00% to 0.05% H2HO. In our case,the acetone-d6 solvent used for preparation (a) contains 0.02%2H2O+H2HO. A calculation of the potential amount of free2H added in the acetone-d6 leads to a ratio of 0.5% to 1% withrespect to MAGF. Consequently, a slow exchange betweenthe free 2H and the OH groups of MAGF could lead to 0.5%to 1% of the MAGF in the –O2H form. This is in goodagreement with the quantity of the O2H form in preparation(a), as estimated from the spectrum of Fig. 2 when it isassumed that the satellites correspond to about 0.5%.

To what extent, then, could the chemical shift isotopeeffect impinge on the accuracy of site-specific 13C isotoperatio measurements performed by 13C NMR? Here, wedemonstrate that very small amounts of residual 2H2O orH2HO in the deuterated solvent (as low as 0.02%) couldlead to an error of about 5‰ to 10‰ in the measured δi

13C.In practice, the small peak of the deuterated isotopologue ofMAGF contributes, in part, to the area of the main peak thatis used to quantify the site-specific 13C isotopomer δi

13C(see Ref. [9] for the description of this determination).Therefore, the error in the calculated δi

13C is minimisedcompared with the actual amount of the deuterated form ofMAGF. As an illustration, we have compared the site-specific 13C distribution in the same glucose analysed as (1)MAGF from preparation (a) (see “Experimental”), containingresidual deuterated species, (2) MAGF from preparation (c)(see “Experimental”) free of exchangeable deuterium and (3)TAMAGF prepared as described in Ref. [15]. The results aresummarised in Table 2. According to the standard deviationof the NMR repeatability of 0.8‰ [15], there is no significantdifference between δi

13C measured on MAGF, preparation (c)

and on TAMAGF. However, an error of up to ∼4‰ isobserved when there is only 0.02% 2H2O+HO2H in theacetone-d6 used for the NMR preparation. It should be notedthat this error does not occur only on C-3, C-5 and C-6(bearing the hydroxyl groups) but is propagated to all the othercarbon positions. This is because the isotopic 13C NMRmethodology does not use an internal reference but rather thepeak molar fraction in combination with the global δ13Cvalue measured by isotope ratio mass spectrometry tocalculate the δi

13C values (see Ref. [9] for details of thiscalculation). Clearly, an error of ∼4‰ is well in excess of therequired performance of the site-specific 13C deviationmeasurement of 1‰.

Strategies to avoid any impact of deuterated impuritieson the accuracy of 13C NMR measurements

When no deuterium is intentionally added to the sample, itis evidently reasonable not to suspect that 2H exchangecould be a problem. In fact, this is not an issue in standardquantitative NMR: it only becomes of concern for isotoperatio measurements and then only in conditions in which aslow exchange of 1H atoms with 2H atoms can occur.

Nevertheless, the range of applications and users ofisotopic 13C NMR is steadily increasing, and it is imperative

Table 1 The β and γ chemical shift isotope effects for the six carbon atoms in the glucose derivative 1,2-O-Isopropylidene)-α-D-glucofuranose(MAGF)

Preparation Δ13C Carbon position

C-1 C-2 C-3 C-4 C-5 C-6

(a) β (ppm) – – 0.103 – 0.107 0.111

(b) β (ppm) – – 0.101 – 0.104 0.113

γ (ppm) – 0.033 – 0.023 0.022 0.032

Effects are calculated as absolute differences between OH and O2 H forms, Δ13 C, expressed in ppm. Numbering is that of glucose conventionalnotation (see Fig. 1)

(a) 200 mg MAGF in 300 μL DMSO and 300 μL acetone-d6, (b) 50 mg MAGF in 700 μL DMSO-d6 and 10 μL 2 H2O

Table 2 Isotopic deviation δi13C (expressed in ‰) for each carbon

position of a sample of glucose measured using three differentpreparations for NMR, and prepared as either the MAGF or theTAMAGF derivative

Sample Carbon position

C-1 C-2 C-3 C-4 C-5 C-6

MAGF (i) −3.9 −5.4 −9.3 −8.6 −15.9 −14.7MAGF (ii) −8.0 −8.6 −6.5 −10.2 −13.6 −10.9TAMAGF (iii) −6.6 −8.7 −8.0 −10.3 −12.4 −11.9

(i) preparation (a) 200 mg MAGF in 300 μL DMSO and 300 μLacetone-d6, (ii) preparation (c) 150 mg MAGF in 300 μL DMSO-d6and 400 μL methanol, (iii) from [15]

Deuterium isotope effect on the accuracy of 13C NMR measurements 1983

that these users are aware that this potential problem existsin those cases where the target molecule contains OH orNH2. The required performance for isotopic 13C NMR isvery high, of the order of per mil (‰), which means thatthe operator must be able to measure a variation of 0.1%of the NMR signal area. Yet, as indicated above, isotopeexchange can introduce an error of about 5‰ to 10‰ inthe measured δi

13C values: such an error is clearlyunacceptable.

Two solutions to circumvent this difficulty are available:first, suppression by appropriate derivatisation of theexchangeable hydrogen positions of the molecule. In theexample of glucose, to overcome this problem, MAGF wasfurther derivatised to yield TAMAGF (Fig. 1), in which thethree OH functions are blocked by an acetyl group.Similarly, the same methodology has been used to studyfructose [15]. In these compounds, no parasite peaks areseen, further supporting the conclusion that they are due tothe isotope shift phenomenon. In addition, it should benoted that the NMR spectrum is further improved both interms of narrower line width and higher signal-to-noiseratio when the hydroxyl groups are suppressed, as inTAMAGF. Such a structure, with fully protected polargroups, allows an easier purification. Secondly, by thejudicious choice of the deuterated solvent, it should bepossible to ensure that the amount of residual 2H2O isbelow the level at which exchange will create significantparasite peaks.

Acknowledgments Alexis Gilbert thanks the Scientific Council ofthe Pays de la Loire Region (France) and the CNRS for a co-fundeddoctoral bursary. We thank Carol Wrigglesworth (Scientific English,Nantes, France) for linguistic correction.

References

1. Billault I, Courant F, Pasquereau L, Derrien S, Robins RJ, NauletN (2007) Anal Chim Acta 593:20–29

2. Martinez I, Aursand M, Erikson U, Singstad TE, Veliyulin E, vander Zwaag C (2003) Trends Food Sci Technol 14:489–498

3. Martin GJ, Akoka S, Martin ML (2006) SNIF-NMR—part 1:Principles. In: Webb GA (ed) Modern magnetic resonance.Springer, Berlin, pp 1629–1639

4. Menditto A, Patriarca M, Magnusson B (2007) Accredit QualAssur 12:45–47

5. Caytan E, Botosoa EP, Silvestre V, Robins RJ, Akoka S, RemaudGS (2007) Anal Chem 79:8266–8269

6. Tenailleau E, Remaud G, Akoka S (2005) Instrum Sci Technol33:391–399

7. Silvestre V, Mboula VM, Jouitteau C, Akoka S, Robins RJ,Remaud GS (2009) J Pharm Biomed 50:336–341

8. Gilbert A (2010) Méthodologies pour l'étude du fractionnementisotopique photosynthétique et post-photosynthétique par RMN 13Cisotopique, Ph.D., thesis, University of Nantes, France, July 8th

9. Remaud GS, Gilbert A, Silvestre V, Botosoa EP, Akoka S, RobinsRJ (2009) Isotopes 2009 Congress; Cluj-Napoca, Romania

10. Caytan E, Cherghaoui Y, Barril C, Jouitteau C, Rabiller C,Remaud GS (2006) Tetrahedron Asymmetry 17:1622–1624

11. Cicchetti E, Silvestre V, Fieber W, Sommer H, Remaud GS,Akoka S, Chaintreau A (2010) Flav Fragr J (in press)

12. Reuben J (1984) J Am Chem Soc 106:6180–618613. Reuben J (1985) J Am Chem Soc 107:1747–175514. Dziembowska T, Hansen PE, Rozwadowski Z (2004) Prog Nucl

Magn Reson Spectrosc 45:1–2915. Gilbert A, Silvestre V, Robins RJ, Remaud GS (2009) Anal Chem

81:8978–898516. Rossmann A, Butzenlechner M, Schmidt HL (1991) Plant Physiol

96:609–61417. Roger O, Lavigne R, Mahmoud M, Buisson C, Onno B, Zhang B-L,

Robins R (2004) J Biol Chem 279:24923–2492818. Robins RJ, Pétavy F, Nemmaoui Y, Ayadi F, Silvestre V, Zhang B-L

(2008) J Biol Chem 283:9704–971219. Bain AD (2003) Prog Nucl Magn Reson Spectrosc 43:63–10320. Bielejewski M, Rachocki A, Luboradzki R, Tritt-Goc J (2008)

Appl Magn Reson 33:431–438

1984 A. Gilbert et al.