Discrimination of Different Hop Varieties Using Headspace Gas Chromatographic Data

Os métodos utilizados para caracterização do lúpulo envolvem a análise de substâncias amargas e do óleo essencial, seguida da comparação da composição química e perfis cromatográficos. No presente trabalho, diferentes variedades e produtos de lúpulo foram discriminados através da análise da composição dos óleos essenciais por cromatografia a gás no modo headspace em fase gasosa, com subsequente tratamento estatístico. Diferenças e similaridades foram determinadas em amostras provenientes dos Estados Unidos, da Nova Zelândia e da Europa, usando-se análise de correlações e “cluster”. Os lúpulos com características de aroma apresentaram baixos níveis de mirceno, enquanto aqueles com características de amargor apresentaram altos teores de β-cariofileno e aloaromadendreno bem como as somas das razões dos marcadores escolhidos, independentemente de estarem sob a forma de “pellets” ou de extratos. Amostras altamente correlacionadas entre si e sem diferenças estatísticas mostram boa possibilidade para substituições entre elas, na produção da cerveja.


Introduction
The composition of hop (Humulus lupulus L.) essential oil has been studied extensively using many different analytical techniques for the separation and identification of the volatile organic compounds.2][3] Loss of relevant volatiles during steam distillation extraction was well documented. 4Although headspace techniques have been recommended to avoid artefacts, the difficulties to obtain quantitative data by using this method has been also pointed out. 5 However, it is well accepted that the headspace technique allows the analysis of a high number of samples in a relative short time allied to rapid and simple sample preparation procedures and it is easily automated.The static headspace gas chromatography, sometimes called direct headspace sampling, is based on the thermodynamic equilibrium between the mixture of volatiles above the sample and the sample itself.When this equilibrium is achieved, generally by closing the sample into a vial which is thermostated in a chosen temperature, an aliquot of the volatile fraction is injected onto the GC column. 6The advance in automatic instrumentation for the injection of headspace samples has allowed a substantial increase in precision of the overall procedure.
In the present work statistical treatment of chromatographic data of hop products was applied for comparison of similarities between different samples.Headspace gas chromatography (HSGC) with automatic injection was used for the analysis of the hop products.The areas of the chromatographic peaks from different samples were compared and treated by correlation and cluster statistical analysis in order to determine significant differences between the chromatographic profiles and also to group similar samples that could be interchanged in beer processing.

Samples
Commercial hop samples were used for this study.Twelve samples were in the form of pellets, one as hexane extract, and nine as CO 2 extracts.The samples from New Zealand were provided by New Zealand Hop (Richmond, New Zealand), the German samples were obtained from John Barth & Sohn GmbH (Wurzburg, Germany) and the American samples were gifts from a Brazilian beer company.Their characteristics are shown in Table 1.
The samples were stored under nitrogen and refrigeration in dark flasks.Before analysis they were left to reach room temperature and were weighed (1.0 g) directly in the headspace vials.The vials were assembled directly into the GC headspace automatic injector.

GC analysis
Gas chromatography was carried out using a Shimadzu 17-A equipment fitted with a HSS-4AE headspace automatic sampler and a C-R7A data processor (Kyoto, Japan).A polyethylene glycol fused silica capillary column from Supelco (Bellefonte, PA, USA) with 50 m x 0.25 mm, i.d., and 0.25 µm of film thickness was used.The oven temperature program started at 50 °C for 10 min, increased to 80 °C at 2 °C min -1 and then to 200 °C at 4 °C min -1 , held for 10 min.The headspace temperature was kept at 110 °C for 30 min previous to the injection of 5 mL of the headspace gas into the column.The injector and detector temperatures were 150 °C and 250 °C, respectively.The carrier gas was hydrogen, at a flow rate of 1.6 mL min -1 and the split ratio was 1:30.Identification of marker compounds was carried out by comparison and spiking with external standards.Myrcene, α-limonene, βcaryophyllene, aloaromadendrene, linalool and αhumulene from Carl Roth GmbH (Karlsruhe, Germany) were used as standards.

Statistical analysis
The areas of the 50 major chromatographic peaks of all samples were considered for the statistical treatment.Cluster Analysis was applied by using a Statgraphics software (Bitstream Inc., Cambridge, MA, USA).Pearson correlation with paired t-test was applied by using a MS EXCEL software (Microsoft Co., Bellevue, WA, USA).

Results and Discussion
The peak areas of the chromatograms obtained by HSGC were used to spot differences between the samples.Figure 1 shows some examples of chromatograms obtained from the samples studied.As it has been commonly applied by other authors, 7 α-humulene was used as reference because it is encountered in all hop samples.Consequently, the area of all chromatographic peaks were then normalized relatively to α-humulene.Reproducibility of data obtained by the HSGC method was very good.The higher coefficient of variation of the relative peak areas was 12.7%, which appears to be quite acceptable in terms of the overall precision normally encountered in headspace chromatographic methods.The use of marker compounds to simplify the analysis of the complex chromatographic data from hop volatiles has been applied to compare the quality of hop varieties. 8In the present work five markers (myrcene, α-limonene, β-caryophyllene, aloaromadendrene and linalool) were identified in all samples.Sample 19 showed highest peak areas for myrcene, α-limonene, β-caryophyllen and aloaromadendrene and these areas were used as denominator to calculate the ratios of those markers in all samples, following the procedure of De Keukeleure. 8inalool showed the highest area in sample 4 and consequently, it was used as denominator for linalool ratios in all samples.The results are shown in Table 2.Both the individual and the sum of the ratios of each sample provided a qualitative and quantitative profile for characterization of the samples, because the predominance of specific ratios are indicative of the hop characteristic showing that the hop product will contribute more to the aroma or to the flavour of the final product.Hop varieties such as Saaz, Perle and Cluster are considered to be some of the most important commercially.Saaz and Perle being typical hop materials while Cluster a hop variety with predominant bitter characteristics.Similar patterns were obtained for samples belonging to the same group of bitter or aromatic products.However, marked differences were detected within groups if samples were in the form of pellets or extract.Sample number 19, which is an extract from a Cluster variety, was completely different from their similar pellet samples, showing higher amounts of all markers except linalaol.However, well known aroma hops such as Saaz and Perle (samples 14 to 16) showed consistently low myrcene content while bitter hops such as Cluster (samples 17 to 20) were richer in βcaryophyllene, aloaromadendrene and also in the sum of the marker ratios (Table 2).
The statistical treatment of the chromatographic data obtained from the volatile profiles allowed to group samples which presented similar characteristics as a whole.A tree diagram with weighted pair group average for the 22 samples shows their hierarchic distribution forming defined groups of aroma and bitter varieties (Figure 2).The pellet samples formed basically two groups (samples 20, 17, 18, 15, 16, 14,11 and 13, 12, 10) while the extracts formed a major group of bitter type hops (samples 19, 22,  5, 4, 21, 7, 6).The New Zealand hops Pacific Gem, Green Bullet and Southern Cross, which show both bitter and aroma characteristics were closer grouped (samples 13, 12 and 10, respectively).Some pair of samples with different characteristics such as 18 and 15 (bitter and aroma, respectively) and 8 and 2 (both aromatic and bitter and bitter, respectively) appears to have mixed properties and may produce similar results in the end product.Sample 3 was grouped together with other Hallertau hop types while sample 9, which is a Sticklebrast variety showed similarities to sample 1, which is a Tettnang Hallertauer type, in the tree diagram.
Because samples 14 to 18 are some of the most commercially important hop products they were selected to determine their correlations with the remaining samples.The statistical comparison of hops of the Saaz variety (14 and 16) and Perle (15) which have strong aroma characteristics, with the other samples allowed the selection of highly correlated samples which could be potentially useful  for substitution in beer formulation.Similarly, confronting Cluster hops (15 and 18), which have bitter characteristics, with the other samples it is possible to choose strongly correlated products in relation to their bitter characteristics (Table 3).
It is important to notice that some samples were not statistically correlated but were statistically similar while others were highly correlated but were significantly different.The high correlated samples which also showed no significant differences, namely 11 with 14, 15, 16 and 18; 15 with 17, 18 and 20; and 16 with 14, 18 and 20, are good possibilities to replace each other in beer formulation.
The establishment of similarities of groups of hop samples with aromatic and bitter characteristics, inde-

Figure 2 .
Figure 2. Tree diagram for 22 variables obtained by statistical treatment of the peak areas of the headspace gas chromatograms of the hop samples.Diagram shows weighted pair-group average and Euclidean distances.

Table 1 .
Characteristics of the hop products