NMR Analysis of Honey
NMR-profiling – The defensive tackle against honey adulteration
Arne Dübecke1, Jane van der Meulen2, Birk Schütz3, Derrick Tanner4, Gudrun Beckh2, Cord Lüllmann2
1Tentamus Global Center of Excellence for Food Fraud, Flughafendamm 9a, 28199 Bremen, Germany
2Quality Services International GmbH, Flughafendamm 9a, 28199 Bremen, Germany
3Bruker BioSpin GmbH, Silberstreifen 4, 76287 Rheinstetten, Germany
4Columbia Food Laboratories Inc., 12423 NE Whitaker Way, Portland, OR 97230

Food fraud awareness increased substantially among authorities, food manufacturers and consumers since the "horse meat scandal" occurred in the UK in 2013. Not only olive oil or expensive spices are targeted by fraudsters. Honey, a highly natural product offered at premium prices, is popular among fraudsters as well, falling third in rank on the list of most frequently adulterated foods, right after milk and olive oil. Phipps strikingly demonstrated the "conundrum" of a strong upward trend in the volume of global honey exports, despite an almost stagnating number of beehives and even decreasing yield per colony. Something is fishy and not all honey imports are what they purport to be.

Types of Fraud
There are several illicit fraud related practices used for honey. Samples may be re-labelled in order to change the actual geographic origin. A reason for that could be the circumvention of anti-dumping duty. A new method of honey screening using Nuclear Magnetic Resonance (NMR) profiling is able to verify geographic and botanic origins, making it useful in tracing transshipment activities, i.e. export of honey from the country of origin to another intermediate country and subsequent relabelling before export to the final destination. The method is described in detail below.

In many Asian countries honey is often harvested too early and unripe. This unripe honey usually lacks the typical taste and odor associated with honey and has far too high a water content, which must be reduced before export. However, the internationally accepted standard for honey of the Codex Alimentarius clearly recommends, that nothing shall be taken out from honey or added to it. The European Honey Directive 2001/110/EC is even more stringent and demands that nothing shall be taken out or added to it. This includes water!

Another common form of adulteration is the treatment of honey with ion exchange resins. These resins can remove residues, e.g. of (prohibited) veterinary drugs and also to lighten very dark honeys, as lighter honey is preferred among consumers and often earns a premium in the market. Before resin treatment, honey must be diluted with water and afterwards the excess water has to be removed again. As we learnt above this is not an approved practice. Furthermore, the resin not only removes veterinary drugs, it also removes a number of natural substances.

The third and most commonly known practice is the addition of cheap sugar syrup. There are several different methods for detection of foreign sugars in honey available. Depending on what kind of syrup has been added, e.g. rice, beet sugar, cane sugar or corn syrup, different analytical methods are used to detect those adulterations.

All these methods share a common downside: they are so-called targeted methods. Targeted methods only look at a single or a small number of very specific parameters, e.g. presence of certain substances or enzymes, which are not naturally found in honey. Of course, the idea behind this approach is good. However, we typically observe that the effectiveness of such targeted methods is highest immediately after introduction to the market. The number of samples found to be adulterated using these methods usually decreases with time. Unfortunately, that does not mean that less adulteration occurs. Rather it shows the successful learning process on the fraudster's side, e.g. they only add syrups that do not contain the marker substances we are targeting with the conventional methods.

NMR-profiling makes the difference.
In cooperation with the world's leading manufacturer of NMR spectrometers, Bruker BioSpin GmbH (Rheinstetten, Germany) and other project partners, QSI GmbH worked on the development of NMR-profiling (HoneyScreener™). This technology performs not only the targeted analyses of a number of parameters, it additionally combines targeted with untargeted detection of many substances at once, i.e. we can "see" sugars, amino acids, organic acids, quality parameters, and many other substances with just a single analysis. What comes out in the end is a "fingerprint" of the honey sample. Of course, a single fingerprint won't do the trick. In addition, you need a database of fingerprints of authentic honey samples to which you can compare your sample in question. Considering the vast diversity of different types of honey this is quite challenging. Of course, this did not keep us from tackling this endeavor!

Since beginning development of the honey database in 2014 it has grown into a very powerful tool. In its present state, the database is unique worldwide. The currently developed database release 2.0 will contain more than 11,000 samples analyzed from a number of countries and monofloral varieties worldwide (Figure 1). All authentic honey samples that were added to the database have been subjected to intensive analytical checks using several conventional methods (e.g. microscopic pollen analysis, carbon isotope ratios (official AOAC method 998.12), foreign enzymes, etc.) to detect adulteration and to assure that only authentic honey fingerprints enter the database. Furthermore, quality criteria must comply with the EU honey directive.

Capabilities of NMR-profiling
NMR-profiling is intended to be used as screening method, e.g. for raw honey before purchase. Especially in conjunction with microscopic pollen analysis, it is very powerful and able to protect buyers (e.g. importers of honey) from unpleasant and unexpected purchasing surprises.
When we evaluate NMR-profiles (fingerprints, Figure 2) we do not just look at the results visually, we also run comprehensive statistical analyses as those can "see" more than one can perceive with eyes alone. To do that, we use established statistical methods like Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). Instead of concentrating on single peaks in the spectrum, we take the entire spectrum and divide it into 500-3,000 little sections (called buckets) which are then fed into the statistical analysis. The PCA is used to reduce the number of variables by finding similarities. This makes it possible to find "hidden" information, e.g. correlations between different parameters, which one could not have discovered by only looking at the NMR-spectra. The LDA is very useful for classification analysis, for example to learn if a sample is of a certain botanic or geographic origin (Figure 3). We can "ask" the algorithm if a sample originates in a specific region like China and the algorithm, also called a model, returns the probability of this sample being from China, helping us to identify samples that were labelled with the wrong geographic or botanical origin.

For the detection of adulteration with syrups and also for treatment with ion exchange resin, we look either for missing peaks that occur in all authentic samples but not in the sample in question, or we search for new peaks, which do not occur in the authentic samples (Figure 4). Care must be taken that new peaks are in fact the result of syrup addition or treatment of honey with resin and not just a specific peak resulting from a rare botanic origin. Due to these circumstances it is of utmost importance to have highly trained experts with a strong background on honey evaluating the NMR-profiles.

In some cases, adulteration takes place through the addition of highly cleaned-up syrups that contain only fructose and glucose in a ratio more or less comparable to that found in authentic honey, but no other residual substances that would typically remain in syrup post production, like oligosaccharides and enzymes. These cases are the most problematic, as there are neither new peaks nor missing peaks in the NMR-profile. Still, we found a way to detect even this type of fraudulent honey dilution. The peaks of fructose and glucose in such adulterated samples appear higher relative to the remaining peaks in the NMR-profile, an effect that can be attributed to the dilution with syrup. Furthermore, evaluation of intensity ratios of certain sections of the spectrum to each other yields valuable indices which are used to gain information on adulteration. Deviations from the typical values of these indices are a very clear sign that something with that honey sample is wrong.

Our NMR-profiling of honey is fully accredited according to ISO 17025, which describes general requirements for testing laboratories. The ISO 17025 is accepted worldwide and compliance with the regulations is a major requirement of almost any client. The accreditation is for the sample preparation and analysis, as well as the applied statistical methods described above.

However, NMR-profiling is not intended to solve every problem. Rather it is a sophisticated screening method ideally used on raw honey before purchase. Like any other analytical method, it's not 100% perfect, and so we recommend using it in conjunction with microscopic pollen analysis as both analyses complement each other very well. In certain cases, additional methods may provide useful information, such as an analysis of the isotopic composition in order to back up NMR-profiling.

In summary, NMR-profiling is a very powerful tool to uncover adulteration of honey. Instead of only targeting single substances, it analyses many substances in honey by combining the targeted with an untargeted approach. The database of authentic honey samples, which is indispensable for accurate results, contains more than 11,000 samples from all major honey exporting countries and relevant floral varieties. As the database continues to grow, the power of the potential analytics improves. NMR-profiling is especially suitable for screening purposes of raw materials and helps protect the buyer of honey from unpleasant surprises.

Scientific Beekeeping
Figure 1: Number of samples per geographic (top) and botanic (bottom) origin. Only those geographic origins with more than 100 samples are shown. Further samples from more than 50 countries are included in the database.
Figure 2: NMR-profile (fingerprint) of an authentic honey from China. The colored areas represent the data of the database samples. The bold black line represents the current sample being analyzed. An authentic honey should mostly follow the red band.
Figure 3: Output of the statistical model for China. In this case the model was "asked" if this sample comes from China. The model compares the sample in question to other samples in the database by performing principal component analysis (PCA) and linear discriminant analysis (LDA). The grey bar on the bottom shows the probability of this sample being from China. In this case the dashed line is clearly passed and the honey's origin in China is confirmed. Equivalent models are used to test botanical origins.
Figure 4: The NMR-profile of adulterated honey often shows "new" peaks that usually do not occur in authentic honey or are missing peaks. Depicted here is a section of the NMR-profile where both cases can be observed: mannose, a sugar that is usually not observed in floral honey, produces a new peak while at other positions in the profile several peaks are completely missing. Both observations clearly point to adulteration.
This article originally appeared in the May 2018 issue of The Scottish Beekeeper Magazine, the journal of the Scottish Beekeepers' Association.