The brewing industry is highly competitive, driving the importance of beer stability. Brewers use both aluminum cans and glass bottles to distribute products; however, a direct comparison of the impact of these two package types on beer stability has yet to be conducted. Here, a non-targeted metabolomics approach was used to characterize changes in the metabolite profile of amber ale (AA) and India pale ale (IPA) packaged in cans and bottles over a 6-month aging period. A strong correlation by package type was observed for AA but not for IPA over all time points. Baseline differences in amino acids (glycine, tyrosine, and asparagine) and esters (isobutyl isobutyrate, 2-methylbutyl butyrate, and ethyl decanoate) were also observed in AA. Hop terpenes (humulene, pinocarvone, and α-calacorene) demonstrated package-dependent changes over time which appear to be influenced by metabolite water solubility. Overall, the results demonstrate that beer metabolites, and thus stability, are significantly impacted by package type.


Beer production is a competitive global industry regardless of the market segment (i.e., international, U.S. macrobrewery, or U.S. craft brewery). In 2020, there were 8,884 breweries in the U.S. (8,764 craft breweries and 120 macro-breweries) that produced 150 million barrels (17.6 billion liters) of beer. (1) Still, in 2020, U.S. beer sales by volume decreased 2.9%, and imported beer accounted for 19.5% of the total sales, signifying increased global competition in the alcohol beverage market. (1) To contend, breweries must use various means to win consumers and retain loyalty, including matching products to consumer preference trends, packaging presentation through container type and label design, and maintaining high quality and flavor stability. A quality beer that builds brand loyalty is one that is free from off-flavors and meets a certain degree of excellence across a complex set of sensory characteristics that are maintained across batches and over time. (2)

Beer is a complex matrix composed of water, ethanol, and thousands of volatile and non-volatile flavor active compounds. Beer flavor is the combination of taste, aroma, and mouthfeel. (3) A beer’s chemical and flavor profile begins to change as soon as it enters the final package. During storage, flavor active compounds undergo chemical reactions that lead to both compound degradation and formation. This process can result in a reduction of favorable sensory attributes and the formation of undesirable ones, contributing to the aging, or staling, of beer. (4) The rate at which these changes occur depends on factors such as dissolved oxygen, temperature, agitation, and raw materials. At a certain point, the flavor profile of a beer is no longer considered “true to brand”. The ability for a beer to maintain its original flavor profile is known as “beer stability” and can be determined through chemical and sensory evaluations. In a competitive global market where the time it takes for product to reach the consumer is increasing, beer stability is an important goal of the brewer and yet remains one of the more difficult quality aspects to control. (5,6)

Many mechanisms of beer aging have been studied and include oxidation reactions, (4,7) Strecker aldehyde (SA) formation, (8) hop volatile and acid degradations, (4,9) and ultraviolet (UV) light effects. (10) However, the majority of published research on beer stability has been conducted on light lagers and a limited group of compounds, such as volatiles and carbonyls. (11) The most studied and notable beer aging marker is trans-2-nonenal, which imparts a paper-like flavor and a drying mouthfeel. (4,5,12) Previous research represents a narrow and dated scope when considering the innovations of new beer styles coming from the modern craft beer industry, the emergence of new hop and barley cultivars, and technological process advancements. Additional research is needed to elucidate the mechanisms of beer stability, identify novel markers of beer aging in diverse beer styles, and characterize the impact of packaging material on beer stability. (5,13)

Glass bottles and cans are the most common single-serve package types breweries use to sell their products to consumers. (14) Over the last decade, there has been a complete shift in container usage in the craft brewing sector. In 2015, the breakdown was 20% cans and 80% bottles. In 2020, the percentages flipped to 80% cans and 20% bottles. (15) The causes of this dramatic shift can be attributed to the recent availability of small canning lines, shifts in consumer activity, and supply chain challenges resulting from the COVID-19 pandemic. While the long-term effects of the pandemic on packaging trends remain to be seen, it is likely that with the increased investment in canning technology across craft breweries, cans will remain the preferred single-serve package type for the foreseeable future.

Although similar in function, cans and bottles are inherently different. Cans are made of aluminum and are coated on the inside with a polymer, often epoxy and acrylic, to protect the liquid from metal taints and the aluminum from corrosion. The can lid is hermetically sealed to the can body, making it an airtight closure. Cans are more susceptible to oxygen pick-up during the packaging process due to the large surface area of the can opening before the lid is attached and the inability to completely purge the can of air via vacuum due to its weak structure. (14) The opaque aluminum material blocks UV light, preventing light-induced reactions that result in an undesirable skunk (“light-struck”) aroma. In contrast, bottles are made of brown, green, or clear glass, although brown is most often used due to its ability to block more UV light. (14,16) Glass bottles undergo air evacuation via vacuum, often resulting in less oxygen pick-up during packing. The cap, or crown, is also lined with a polymer similar to the coating of cans and is permeable at the glass–liner interface, allowing air to ingress over time. (17) The differences between cans and bottles affect the risk potential for different types of aging reactions, adding container type to the list of variables that affect flavor stability in beer.

We hypothesized that package type will influence beer stability and that this impact will be beer style-dependent. To test this hypothesis, we utilized a non-targeted metabolomics approach to explore the effects of packaging type on beer chemical stability in two craft-relevant, non-light lager styles. The use of a non-targeted analytical approach allowed for the novel discovery of metabolites important to beer stability. Our focus on styles beyond light lagers and the inclusion of various packaging types extends our knowledge of beer stability and enables new insights into flavor stability mechanisms. The results of this study are relevant to modern breweries and beer styles and provide valuable information to enable scientifically backed decision-making around packaging and package-type best practices.

Materials and Methods

Brewing Parameters and Storage

Sample beer was brewed at New Belgium Brewing Company (Fort Collins, CO). One batch each of an amber ale (AA) (5.2% ABV, 22 IBU) and India pale ale (IPA) (7.0% ABV, 50 IBU), from the same brite tank was packaged into both bottles and cans. Packaging of the two container types occurred on the same day for each batch, and packaging of the two batches occurred within the same week. Both canned and bottled samples were flash pasteurized as a final stabilization step in the packaging process. The IPA was packaged into 12 oz. aluminum cans and brown glass bottles, and the AA was packaged into 16 oz. aluminum cans and 12 oz. brown glass bottles. Post-packaging, the batches were tested in-house and passed analytical and sensory quality parameters, indicating an acceptable representation of each product. Total package oxygen (TPO) was measured by Pentair Haffmans Automatic Inpack TPO/CO2 meter during the packaging run. Measurements were taken throughout the run and then averaged. The averages passed brewery quality control (QC) specifications and were as follows for each of the four treatments: AA cans (41.3 ppb TPO), AA bottles (85.4 ppb TPO), IPA cans (76.5 ppb TPO), and IPA bottles (104 ppb TPO). These TPO values fall within typical industry standards of 0–150 ppb TPO as an acceptable in-package QC range.

Immediately after packaging, samples were stored under cold conditions (3 °C) at the brewery. Due to logistical constraints, the time between package day and the week zero, or baseline, sample was 14 days for AA and 6 days for IPA. Once all samples were packaged, they were transferred to Colorado State University (CSU), where they were stored under cold conditions (3 °C) for the first 30 days, then stored at room temperature (20 °C) for 150 days, for a total of six months storage time. Storage conditions were established to mimic typical market storage conditions. Shelf life of both brands were previously determined by a trained sensory panel (trained and operated by New Belgium Brewing) to be 5 months.

Sampling and Sample Preparation

Throughout the 6-month storage period, samples were collected biweekly, starting with the day samples were transferred to CSU (i.e., week zero, or baseline), resulting in a total of 13 time points. At each sampling time, three unique packaged samples (n = 3) in each treatment were randomly selected from storage, resulting in a total of 156 samples over the course of the storage period. For each sample, multiple aliquots were collected in 2 mL glass vials and stored at −80 °C until chemical analysis. At the time of chemical analysis, samples were randomized, thawed, and conditioned to room temperature (20 °C) over a 1 h period, at which point the samples were degassed by sonication.

GC–MS Analysis

A sample clean-up procedure was performed prior to derivatization. This process precipitated out large polysaccharides to prevent interference with metabolite detection. 200 μL degassed beer, 400 μL methanol, and 400 μL acetonitrile were added to a glass vial and agitated for 1 h at 20 °C, followed by a 20 h incubation at 3 °C, and then centrifugation at 3500 rpm. A pooled QC sample comprised of 22 μL from each sample was created with the supernatant from this step. 75 μL of the supernatant from each sample and pooled QC was transferred to a new glass vial each and dried down for 25 min using nitrogen gas. Vials were capped and then stored at −80 °C until derivatization.

Samples were derivatized by methoxyamine HCl (25 mg/mL pyridine) and MSTFA+1% TMCS. The glass vials containing the dried-down supernatants were brought to room temperature, and 50 μL methoxyamine HCl (25 mg/mL pyridine) was added to resuspend the contents. Vials were incubated at 60 °C for 45 min then sonicated for 10 min before a second incubation at 60 °C for 45 min. 50 μL MSTFA + 1% TMCS was added, vortexed, and incubated for a final round at 60 °C for 40 min. Vials were brought to room temperature over 10 min.

A Clarus 690 gas chromatography system (PerkinElmer Waltham, MA, USA) coupled to a PerkinElmer Clarus SQ 8T mass detector was used for the separation and detection of small semi-volatile and non-volatile molecules. Separation was conducted with a TG-5MS column (Thermo Scientific, 30 m × 0.25 mm × 0.25 mm). One microliter of the derivatized sample was injected at a 1:12 split ratio and 1.0 mL/min helium gas flow. Samples were injected in a randomized order with a pooled QC injected between every six samples. The oven profile consisted of an 80 °C hold for 30 s, ramping 15 °C/min to 330 °C, with an 8 min hold at the end of the run. Masses between 50 and 620 m/z were scanned at 4 scans/s after electron impact ionization operating at 70 eV. The injector temperature was held at 285 °C, the transfer line was held at 280 °C, and the source was held at 260 °C.

For all samples, 2 mL of degassed beer was transferred to a 20 mL headspace (HS) vial and immediately capped with a crimper. A pooled QC sample comprised of 500 μL of each sample was prepared using the degassed room temperature samples. Samples were analyzed in a randomized order, and a pooled QC was injected every six samples. Samples were stored at 3 °C until loaded in the autosampler.

For headspace analysis, separation was conducted with an Elite-624Sil MS column (PerkinElmer, 30 m × 0.25 mm × 1.4 mm). The TurboMatrix 40 Trap headspace sampler (PerkinElmer) was used as an HS sampler. Prior to injection, samples were heated to 80 °C for 20 min, followed by preconcentration for one cycle in a TurboMatrix air monitoring trap (PerkinElmer). Absorption of compounds to the trap was performed at 25 °C at a vial pressure of 40 psi. Vaporization was performed at 20 psi and 260 °C for 0.5 min. The HS needle was held at 120 °C, and the HS transfer line was held at 140 °C. Samples were injected at a 1:10 split ratio and 1.0 mL/min helium gas flow with the column pressure set to 23 psi. The oven profile consisted of a 35 °C hold for 5 min, ramping 6 °C/min to 245 °C. Masses between 35 and 350 m/z were scanned at 4 scans/s after electron impact ionization operating at 70 eV. The injector temperature was held at 180 °C, the transfer line was held at 260 °C, and the source was held at 240 °C.

Metabolite Annotation and Statistical Analysis

The open source XCMS software was used to define a matrix of molecular features as previously described. (18) The RAMClustR package operating within the R Programming (version 4.0.3) environment was used to cluster co-varying and co-eluting features. (19) Metabolite annotations were performed by spectral searching against the NIST, GOLM, and an in-house proprietary database in the program RamSearch (Colorado State University, CO, USA). (20) Data from each analysis [i.e., gas chromatography–mass spectrometry (GC–MS) and headspace GC–MS (HS-GC–MS)] was processed independently. The resulting datasets were then z-score-transformed and combined into a final data matrix and used for multivariate (MVA) statistical analyses. The SIMCA software (Umetrics, Version 17) was used to perform principal component analysis (PCA) and partial least-squares discriminant analysis (PLS-DA), with data scaled by unit variance. Variable importance of the projection (VIP) scores from model biplots was used to identify metabolites of interest. Those metabolites with a VIP score ≥2 were flagged for further statistical analysis. Multivariate analysis of variance (MANOVA), performed in R Programming using the manova() function of the mvnormtest package, was also used to screen and identify metabolites of interest (P ≤ 0.05).

Univariate analysis (UVA) was then used to explore the final group of metabolites of interest. Assumptions of normality were checked, and log or square root transformations were performed if necessary. A linear model was created, and model fitness was assessed with the multiple R2 value. Metabolites with a linear model fit of R2 ≥ 0.900 were considered adequate. The change over time and the estimated marginal means at specific time points were explored using the emmeans() and emtrends() function of the emmeans package for R Programming, (21) where significance was determined at the 95% confidence interval and P ≤ 0.05.

Results and Discussion

Beer Type Was the Largest Source of Chemical Variation across All Samples

Non-targeted metabolite profiling is a powerful tool that can be used to detect global metabolite variations in a large dataset. In the present study, this approach was used for all sample treatments (i.e., AA can, AA bottle, IPA can, and IPA bottles) across 13 timepoints (n = 156), resulting in the detection of 351 molecular features across both analytical platforms (Supporting Information). Of those molecular features, 73 total metabolites were annotated, 46 by GC–MS detection and 27 by HS-GC–MS detection (Table S1). The analytical platforms utilized in this study resulted in the detection of volatile and small non-volatile polar compounds, including amino acids, terpenes, esters, alcohols, carbohydrates, and carbonyls. The metabolites and their relative abundances from both detection platforms were combined into one data matrix and normalized about the mean (i.e., z-score) prior to MVA. An unsupervised PCA was performed on the z-scored relative abundance values for the 73 annotated metabolites in all samples after outliers were removed (n = 144) to determine the source of variation across the entire dataset (Figure 1). Outliers were defined as datapoints with a Hotelling’s T2 Crit >95%. The PCA model explains 74.5% of the total variation across the first two components, where principal component (PC) 1 explains 69.9% of the variation and PC2 explains 4.6% of the variation. The model is well fit (R2 = 0.867) and clearly separates AA and IPA across PC1, indicating style as the main source of variation across all samples. This is to be expected with AA and IPA being unique and vastly different beer styles. Amber ales are a malt forward style made with a low-to-medium level of hops utilized in the recipe. IPAs have a prominent hop profile due to the higher hopping rates in their recipe formulation. Thus, it is expected that metabolite profiles of different styles should be the greatest source of variation, regardless of storage time or package type. The PCA model supports this expectation and thus serves as a positive check on our dataset.

Figure 1. PCA of combined GC–MS and HS-GC–MS datasets. The combined data matrix used in the formation of this model was normalized about the mean (z-score). The model is separated by style across PC1, with brown circles representing AA samples and green diamonds representing IPA samples.

Impact of Package Type Is Dependent upon Beer Style

To further investigate package type differences, a supervised PLS-DA was applied only to the AA samples with the factors defined for package type (cans and bottles) (Figure 2a). The model fit is good and predictive (R2 = 0.981, Q2 = 0.964). This provides evidence that there is significant metabolite variation between AA cans and AA bottles. The loadings from the biplot show which metabolites are driving the variation in the data.

Figure 2. Biplots of PLS-DA models for AA and IPA by package type. (a) PLS-DA model for AA with brown circles representing bottled samples and gray circles representing canned samples. Model is separated by package type across PC1, exhibiting package type predictability. Metabolite loadings are indicated by blue triangles. (b) PLS-DA model for IPA shows a lack of separation (overlap of can and bottle scores) and predictability by packaging type.

A PLS-DA model was also built with the data from IPA cans and IPA bottles (Figure 2b). While a general grouping of the IPA bottle samples on the left and IPA can samples on the right can be observed, the model fit and predictive power (R2 = 0.667 and Q2 = 0.115, respectively) are weak, indicating a lack of significant metabolite variation due to package type. Taken together, these results indicate that the impact of package type is style-dependent.

The VIP value for each metabolite in the model represents its importance in driving the observed variation. For both PLS-DA models, metabolites with a VIP ≥2 were flagged as metabolites of interest for further statistical investigation. Additionally, a MANOVA test was conducted on the two datasets, and the metabolites with significance (P ≤ 0.05) in the interaction between storage week number and package type were also flagged as metabolites of interest. In total, 17 metabolites consisting of amino acids, ester, terpene, alcohol, and carbonyl chemical classes were flagged for further investigation. A complete list of the metabolites of interest, their documented sensory attributes, and the methods used to identify them can be found in Table 1.

Table 1. Metabolites of Interest as Defined by Either a VIP Score ≥2 or a MANOVA P-Value ≤0.05

classsubclassmetabolitesensory attributedetectionVIPaVIPbP-valuec
carboxylic acidamino acidsglycineNAdGC–MS2.000.760.81
 carboxylic acid esterethyl acetatenail polish remover, solvent, fruity, sweetHS-GC–MS0.221.020.02
  isobutyl isobutyrategrape skin, pineapple, tropicalHS-GC–MS2.180.120.71
lipidfatty acid esterethyl decanoatecaprylic, soapy, esteryHS-GC–MS0.591.42<0.001
  ethyl octanoateapple, sweet, fruity, sour appleHS-GC–MS0.590.43<0.001
  ethyl hexanoateapple, anise seed, citrus, solventHS-GC–MS0.560.590.02
  2-methylbutyl butyratefruity, pear, apricot, tropical, spicy, appleHS-GC–MS2.130.870.40
  β-myrcenespicy, citrus, resinous, piney, lemon, woodyHS-GC–MS0.792.460.74
  β-pinenewoody, green, resinous, dryHS-GC–MS0.592.450.25
 sesquiterpenehumulenespicy, herbal, grassy, woody, cloveHS-GC–MS1.711.76<0.001
  α-calacorenecitrus, spicy, woodyHS-GC–MS0.662.09<0.001
organooxygenalcoholisobutanolmalty, solventHS-GC–MS0.220.540.01
 carbonyl2-undecanonevarnish, bitter, green plants, geranium, fruity, citrusHS-GC–MS0.581.590.005

a VIP scores from the PLS-DA model of AA samples.

b VIP scores from the PLS-DA model of IPA samples.

c P-values from the analysis of variance (ANOVA) and the interaction between the package type and aging week number.

d Not applicable (NA) due to a lack of sensory information in the literature.

Figure 3 represents the relationship between the package type and the relative abundance of the 17 selected metabolites of interest over the aging time. Each rectangle represents the average abundance for a metabolite (n = 3) at each aging timepoint. Most metabolites are lower in relative abundance in the AA samples than in the IPA samples. Interestingly, ethyl decanoate, ethyl octanoate, and isobutanol follow an opposite trend. Furthermore, there is a metabolite group consisting of myo-inositol, asparagine, and glycine that were lower in AA bottles than in AA cans. Uniquely observed in the IPA bottles, the hop acids humulene, β-myrcene, and β-pinene clearly decreased in abundance over the aging time, while another hop acid, α-calacorene, increased with aging.

Figure 3. Heatmap visualization of relationships between treatment, metabolite, and time. The figure was created with the averaged z-score normalized relative abundances (n = 3) of the metabolites of interest for each sample treatment and biweekly sample time point (n = 13).

Baseline Differences in Metabolite Abundance Explain Package Type Variation in AA

Of the 17 metabolites of interest, eight had significant (P ≤ 0.05) baseline differences between the package type in the AA samples and six in the IPA samples (Table 1Figure 4). Six out of eight of the metabolites with significant baseline differences in AA were also flagged from the VIP scores of the PLS-DA model. This suggests that baseline differences in metabolite abundance explain, in part, the separation observed between the cans and bottles in the AA PLS-DA model.

Figure 4. Generalized baseline differences in AA samples by chemical class. (a) Proportion graph of chemical classes with baseline (week 0) differences in AA samples, (b) proportion graph of chemical classes with baseline (week 0) differences in IPA samples, (c) amino acids (Gly = glycine, Try = tyrosine, and Asp = asparagine) with significant (*** indicates P < 0.001) baseline differences between cans and bottles in AA samples, and (d) esters (Isobut = isobutyl isobutyrate, 2MB = 2-methylbutyl butyrate, Edec = ethyl decanoate) with significant (*** indicates P < 0.001) baseline differences between cans and bottles in AA samples.

Two main chemical classes, esters and amino acids, stand out from the list of eight metabolites. Isobutyl isobutyrate (Isobut) and 2-methylbutyl butyrate (2MB) are two esters for which the baseline abundance was observed to be significantly higher in AA bottles than AA cans. Esters are highly volatile and thus readily move from beer to the atmosphere and are susceptible to oxidation. (4) The larger air-exposed surface area in a can prior to sealing could result in increased volatilization of esters from beer during packaging. This would result in a lower baseline abundance of these compounds in the can as compared to the bottle. The greater air-exposed surface area could also result in increased oxidation reactions in cans as compared to bottles from the initial total packaged oxygen (TPO). It has been shown that esters are readily oxidized and that these reactions occur after packaging once the ground state oxygen that makes up the TPO becomes reactive oxygen species (ROS). (4,22) There is a lag phase between when TPO becomes ROS and begins oxidative deterioration. The length of the lag phase varies with different beer matrixes and their endogenous antioxidant properties. (23) It appears the lag phase occurred within the 14 day time gap between when AA samples were packaged and when the baseline samples were collected in the lab. The difference in antioxidant properties between the beer styles may explain the lack of observed baseline differences in the IPA samples. For example, hops contain hundreds of polyphenols that have well-documented antioxidative properties. (6,24) Because IPA styles utilize high hopping rates, the endogenous antioxidant activity of the IPA matrix protects against oxidation compared to the lower-hopped AA samples.

Esters provide important flavor and aroma characteristics in beer. Each style or brand has its own unique group of esters that make up the flavor profile along with other flavor active compounds. A reduction in esters will result in an overall dampening of the flavor, whereas the formation of esters will impart unintended aromas. In both scenarios, the overall balance of the beer is affected. (13,25) Ultimately, it is up to the brewer to decide which esters are important to each of their brands and if a change in a specific ester’s abundance will negatively affect the organoleptic properties of that brand.

The amino acids tyrosine (Tyr), glycine (Gly), and asparagine (Asp) were observed to be significantly lower in abundance in AA bottles than cans at baseline. Interestingly, no significant difference in these metabolites was observed in the IPA samples. It is well understood that amino acids will bind to silica glass, and this is often a source of analytical error when preparing samples for amino acid analysis in a lab setting. (26) Although not previously documented, it is possible that amino acids in beer adsorb to the inside of a glass bottle, accounting for the baseline differences between package types observed for the AA beer style. To understand the style-dependent effect of baseline differences of amino acids, we consider the differences in hop usage between AA and IPA. The antioxidant properties of hops are the result of hundreds of polyphenol compounds. (24) It has been extensively documented that these polyphenols will bind to proteins, peptides, and therefore amino acids. (6,27) The higher hopping rates used in the production of IPAs result in increased levels of polyphenols in the solution that could bind the amino acids in the beer matrix. Furthermore, the amino acids in beer primarily originate from malt, and thus higher concentrations are seen in the IPA sample because of its higher finishing gravity. If amino acids are bound to polyphenols, the adsorption effect would be impeded, which could reflect the lack of baseline difference observed between package types in the IPA style.

Recent work has shown that free amino acids in beer will react with Maillard compounds formed during mashing and undergo the Strecker degradation reaction. (4,22,28) The resulting Strecker aldehye (SA) products are associated with negative staling flavors and aromas. (4,22) Each amino acid will result in a unique SA, (28) and their sensory impacts continue to be explored. In beers with lower hopping rates where amino acid – polyphenol binding is minimal, canned products could be at a higher risk for SA formation due to the higher abundance of amino acids at baseline from a lack of adsorption effects. This could ultimately lead to sensory differences of the same batch of beer packaged into both bottles and cans over time.

Taken together, these results demonstrate that baseline differences are a major variable explaining the predictability of package type in AA samples, and therefore, the discussion has focused on the metabolites with significant baseline differences in this style. Targeted investigations on the mechanisms driving baseline differences warrant continued exploration.

Package Type Impacts Metabolite Variation during Storage

The P-values reported from ANOVA (Table 1) reveal significant changes in metabolite abundance over time that are dependent on the package type. The estimated marginal means of linear trends was performed once assumptions of normality were met and a linear model was fitted. The analysis produced a 95% confidence interval, indicating a significant change over time for a metabolite in each sample treatment (Table S2).

Of the 17 metabolites of interest, 10 were found to be significant from the estimated marginal means of linear trend analysis. Six metabolites significantly changed in abundance over the 24-week storage period in IPA bottles, five in IPA cans, four in AA bottles, and ten in AA cans (Figure 5a). All four of the metabolites identified in AA bottles were also identified in AA cans. The additional six metabolites in AA cans were tyrosine, 2MB, myo-inositol, ethyl acetate, ethyl octanoate, and ethyl hexanoate. It is important to note the later three metabolites did not have an acceptable model fit, and therefore, conclusions regarding their influence on the variation in the data are limited. Still, there are six metabolites that significantly changed in abundance over aging time uniquely in AA cans as compared to bottles. This result suggests that there was a reduction in chemical stability in AA packaged in cans relative to bottles. Whether these compounds have an impact on the organoleptic properties of the beer is outside the scope of this study but warrants further investigation.

Figure 5. Generalized significant changes over time by chemical class and liner modeling of important terpene metabolites. (a) Proportion charts for each sample treatment and the chemical classes significantly (95% confidence interval) changing over time of the 17 identified metabolites of interest. (b) Linear modeling of relative abundance (RA) of four important terpenes (humulene, β-myrcene, α-calacorene, and pinocarvone) significantly (P ≤ 0.05) changing over time. Each terpene linear model graphs the four sample treatments, with orange representing AA, green IPA, dashed line/circle symbol bottle, and a solid line/triangle symbol cans. Each time point symbol represents the mean (n = 3) with standard error bars.

Terpene Water Solubility Impacts Relative Abundance Changes during Storage

Four of the metabolites of interests are hop terpenes: humulene, β-myrcene, α-calacorene, pinocarvone (Figure 5b). These terpenes exhibit a decrease in relative abundance during the storage period, apart from α-calacorene, which increased. Linear modeling of these four terpenes over time reveals how each metabolite changes depending on the package type and style. Humulene, a hop volatile with sensory descriptors of earthy and herbal, significantly decreased in abundance in both AA and IPA and is dependent on the package type (P < 0.001). The linear models for both beer styles show that humulene decreased with a greater magnitude in bottles as compared to cans. Myrcene, a hop volatile with sensory descriptors of hoppy and freshly herbaceous, also significantly decreased in abundance over time in both cans and bottles. Unlike humulene, package type was not significant (P = 0.74), and the rate of decrease was statistically the same for IPA and AA bottles and cans. Pinocarvone, described as minty, demonstrated a significant decrease (P = 0.001) in relative abundance over time in IPA cans only.

These results are consistent with previous work that demonstrated flavor scalping of beer terpenes by can and bottle package material dependent upon water solubility. (9,12) Flavor scalping in packaged food and beverages is the movement of flavor active compounds from the product into the packaging material resulting in a decreased sensory intensity. (29) Temperature, alcohol content, pH, and metabolite water solubility can affect the degree to which a metabolite is scalped by its packaging material. In packaged beer, the polymer lining of a can and the polymer-lined underside of a bottle cap are known to scalp terpenes. (9,30) Beer is a relatively polar liquid consisting mostly of water, and the polymer linings of cans and bottle caps are non-polar. Metabolites with low water solubility will more readily move out of the beer and into the packaging lining. The magnitude of this effect is greater as water solubility decreases. (9,12)

The results of the present study demonstrate these principles. The water solubilities of humulene, myrcene, and pinocarvone are 0.011 g/L, 0.077 g/L, and 0.62 g/L, respectively. Humulene, with the lowest water solubility, is scalped to the greatest degree, and myrcene to a lesser degree. (9,30) With the highest water solubility, pinocarvone will remain in beer more than humulene and myrcene, which may explain why a significant decrease was only observed in IPA cans. Beer is in constant contact with the can liner, which increases the likelihood of scalping pinocarvone. This is in contrast to bottles, where highly water-soluble pinocarvone will not as readily enter the headspace of a bottle and therefore come into less contact with the cap liner in bottled IPAs. The muted scalping effect of pinocarvone in the AA samples is likely a result of the lower starting relative abundance of this compound in AA style beer as compared to IPA.

The observed higher scalping rate of humulene in bottles versus cans could be explained by a mechanism of equilibrium. In cans, beer is in constant contact with the can liner, providing an opportunity for equilibrium to be reached. However, in bottles, once the volatile humulene in the headspace is absorbed by the cap liner, there is limited contact with the beer, and thus the humulene remains “trapped” in the liner. Furthermore, oxygen ingress through bottle crowns, a well-studied phenomenon in bottled beer, (31) could increase the oxidation reactions occurring in IPA bottles. It could also be presumed that, in addition to oxygen ingress, gas in the headspace will egress to the ambient environment, providing more volume for volatiles with low water solubilities to move from liquid beer into the headspace further exacerbating the diminishing of those compounds (e.g., humulene).

α-calacorene, a compound known to impart spicy and herbal aromas, was the only terpene to increase in abundance during storage. (32) This increase is package-type dependent (P < 0.001) and is observed in both AA and IPA. The rate of change is greater in cans for both AA and IPA, suggesting that package type has a greater impact than style on the observed changes. The increase in α-calacorene could be explained by its formation through oxidation or enzymatic release, which has been previously demonstrated for other terpenes; (33) however, further work is needed to fully understand the mechanisms causing the increase of this terpene.

It is not surprising that the terpenes detected in this study exhibit varying behaviors during aging due to their unique chemical properties, such as water solubility. The implications for flavor are greater for IPAs due to the importance of hop volatiles in IPA styles; however, regardless of style, any change in the abundance of flavor active compounds could affect the chemical balance and impact the original sensory profile.

In the present study, the use of a non-targeted analytical approach allowed for novel discoveries of metabolites and potential mechanisms important to beer stability in different beer styles and packaging types. The impact of package type was dependent on style, where AA beer demonstrated significant differences between cans and bottles that were not observed in IPA beer. Furthermore, baseline differences in metabolite abundance between cans and bottles were the primary driver in the overall metabolite variation in AA beer samples, regardless of aging time. The proposed mechanisms are oxidation, volatilization during packaging, and adsorption to glass packaging material. Although the impact of package type in IPA was weak, changes in metabolite abundance over aging time, particularly in terpenes, provide valuable insight into flavor-specific aging mechanisms in IPA beer. Aligning with previous work, these changes vary based on the terpene chemical properties (e.g., water solubility), which will determine their interaction with the packaging material, specifically the can and bottle crown liners.

Taken together, the results of this study do not support the conclusion of a general best package for all beer styles but rather indicate that the effects of package type are dependent on beer style. Continued work to define the mechanisms driving the effects of package type on beer stability is warranted and would require integration with sensory outcomes. Ultimately, research on beer stability and packaging should provide relevant knowledge, so brewers may make scientifically backed packaging decisions and shelf life determinations.

Data Availability

GC–MS and HS-GC–MS data have been deposited to the MassIVE database (DOI: 10.25345/C5959CC5B) with the identifier MSV000090380. The complete dataset can be accessed here: https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp.

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsfoodscitech.2c00351.

  • List of annotated metabolites from GC-MS (Table S1) and HS-GS-MS (Table S2) including the linear models of metabolite abundance over aging time PDF)
  • Normalized spectral abundance and z-score for annotated compounds from GC-MS and HS-GC-MS analysis (XLSX)

Terms and Conditions

Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.

Author Information

  • Corresponding Author
  • Authors
    • Kathryn Fromuth – Department of Horticulture and Landscape Architecture, Colorado State University, Fort Collins 80523, Colorado. United States
    • Jacqueline M. Chaparro – Department of Horticulture and Landscape Architecture, Colorado State University, Fort Collins 80523, Colorado. United States
    • Dana Sedin – New Belgium Brewing Company, Fort Collins, Colorado 80523, United States
    • Charlene Van Buiten – Department of Food Science and Human Nutrition, Colorado State University, Fort Collins 80523, Colorado. United States


We are grateful to New Belgium Brewing Company for their donation of beer samples for this study as well as for expert guidance throughout the project from inception to completion. We would also like to thank the Graybill Statistics and Data Science Laboratory at Colorado State University and Dr. Zachary Weller for their consultation and support in designing the appropriate statistical analysis approach to answer the project’s research questions.

Note Added After ASAP Publication

This paper was published ASAP on March 13, 2023, with an error in the Introduction. The corrected version was reposted on March 29, 2023.


ABValcohol by volume
IBUinternational bittering units
AAamber ale
IPAIndia pale ale
QCquality control
MSTFA TMCSN-methyl-N-(trimethylsilyl) trifluoroacetamide
TPOtotal packaged oxygen
CO2carbon dioxide
GC–MSgas chromatography–mass spectrometry
HS-GC–MSheadspace gas chromatography–mass spectrometry
PCAprincipal component analysis
PCprincipal component
PLS-DApartial least-squares discriminant analysis
MVAmultivariate analysis
UVAunivariate analysis
MANOVAmultivariate analysis of variance
VIPvariable importance projection
NAnot applicable
Isobutisobutyl isobutyrate
2MB2-methylbutyl butyrate
Edecethyl decanoate
ROSreactive oxygen species;
SAStrecker aldehydes
ANOVAanalysis of variation


This article references 33 other publications.

  1. 1(a) National Beer Sales & Production Data. Brewers Association, 2020. https://www.brewersassociation.org/statistics-and-data/national-beer-stats/ (accessed September 9, 2021).Google ScholarJohn Dunham & Associates. The U.S. Beer Industry’s Economic Contribution in 2020: Analysis, Methodology, and Documentation; Beer Institute, 2021. https://www.beerinstitute.org/wp-content/uploads/2021/05/2020-Beer-Serves-America-Report.pdf (accessed Sep/19/2021).Google Scholar
  2. 2 Beer flavour and sensory assessment. In Brewing: Science and practice. Briggs, D. E., Boulton, C. A., Brookes, P. A., Stevens, R., Eds. Woodhead Publishing Series in Food Science, Technology and Nutrition; Brewing Woodhead Publishing, 2004; pp 716– 758.[Crossref], Google Scholar
  3. 3 Best practices guide to quality craft beer: Delivering optimal flavor to the consumer, Association, B., Ed.; Brewers Association www.brewersassociation.org.Google Scholar
  4. 4Vanderhaegen, B.; Neven, H.; Verachtert, H.; Derdelinckx, G. The chemistry of beer aging – a critical review. Food Chem. 200695, 357– 381,  DOI: 10.1016/j.foodchem.2005.01.006 [Crossref], [CAS], Google Scholar
  5. 5Bamforth, C. W. 125th Anniversary Review: The Non-Biological Instability of Beer. J. Inst. Brew. 2011117, 488– 497,  DOI: 10.1002/j.2050-0416.2011.tb00496.x [Crossref], [CAS], Google Scholar
  6. 6Aron, P. M.; Shellhammer, T. H. A Discussion of Polyphenols in Beer Physical and Flavour Stability. J. Inst. Brew. 2010116, 369– 380,  DOI: 10.1002/j.2050-0416.2010.tb00788.x [Crossref], [CAS], Google Scholar
  7. 7Kuchel, L.; Brody, A. L.; Wicker, L. Oxygen and its reactions in beer. Packag. Technol. Sci. 200619, 25– 32,  DOI: 10.1002/pts.705 [Crossref], [CAS], Google Scholar
  8. 8Ferreira, I.; Guido, L. Impact of Wort Amino Acids on Beer Flavour: A Review. Fermentation 20184, 23,  DOI: 10.3390/fermentation4020023 [Crossref], [CAS], Google Scholar
  9. 9Wietstock, P. C.; Glattfelder, R.; Garbe, L.-A.; Methner, F.-J. Characterization of the Migration of Hop Volatiles into Different Crown Cork Liner Polymers and Can Coatings. J. Agric. Food Chem. 201664, 2737– 2745,  DOI: 10.1021/acs.jafc.6b00031 [ACS Full Text ], [CAS], Google Scholar
  10. 10(a) Stewart, G. G.; Priest, F. G. Beer Shelf Life and Stability. In The stability and shelf life of food; Subramaniam, P., Ed.; In Woodhead Publishing Series in Food Science, Technology and Nutrition, 2nd ed.; Woodhead Publishing, 2016; pp 293– 309.[Crossref], Google Scholar(b) Nakayama, T. O. M.; Fly, W. H. Volatile Compounds from the Photodegradation of Hop Acids. Proceedings. Annual meeting – American Society of Brewing Chemists 196826, 198– 202,  DOI: 10.1080/00960845.1968.12006369 [Crossref], Google Scholar
  11. 11Hashimoto, N.; Eshima, T. Composition and Pathway of Formation of Stale Aldehydes in Bottled Beer. J. Am. Soc. Brew. Chem. 197735, 145– 150,  DOI: 10.1094/asbcj-35-0145 [Crossref], [CAS], Google Scholar
  12. 12Jamieson, A. M.; Van Gheluwe, J. E. A. Identifcation of a Compound Responsible for Cardboard Flavor in Beer. Proceedings. Annual meeting – American Society of Brewing Chemists 197028, 192– 197,  DOI: 10.1080/00960845.1970.12006980 [Crossref], Google Scholar
  13. 13Vanderhaegen, B.; Delvaux, F.; Daenen, L.; Verachtert, H.; Delvaux, F. R. Aging characteristics of different beer types. Food Chem. 2007103, 404– 412,  DOI: 10.1016/j.foodchem.2006.07.062 [Crossref], [CAS], Google Scholar
  14. 14 Packaging. In Brewing: Science and practice; Briggs, D. E., Boulton, C. A., Brookes, P. A., Stevens, R., Eds.; Woodhead Publishing Series in Food Science, Technology and Nutrition,Brewing; Woodhead Publishing, 2004; pp 759– 811.[Crossref], Google Scholar
  15. 15Watson, B. Craft Beer Packaging Trends Recap, 2020. Brewers Association. https://www.brewersassociation.org/insights/craft-beer-packaging-trends-recap-2020/ (accessed September 9, 2021).Google Scholar(b) Pitts, E. R.; Witrick, K. Brewery Packaging in a Post-COVID Economy within the United States. Beverages 20217, 14,  DOI: 10.3390/beverages7010014 [Crossref], Google Scholar
  16. 16Lewis, M. J.; Young, T. W. Beer packaging and dispense. In Brewing, 2nd ed.; Aspen Publishers, Inc., 2001; pp 351– 368. DOI: 10.1007/978-1-4615-0729-1_20 [Crossref], Google Scholar
  17. 17Barth, R. The chemistry of beer styles. In The chemistry of beer: the science in the suds; John Wiley & Sons, Inc., 2013; pp 211– 224.[Crossref], Google Scholar
  18. 18(a) Yao, L.; Sheflin, A. M.; Broeckling, C. D.; Prenni, J. E. Data Processing for GC-MS- and LC-MS-Based Untargeted Metabolomics. In High-Throughput Metabolomics; Springer: New York, 2019; pp 287– 299.[Crossref], Google Scholar(b) Broeckling, C. D.; Afsar, F. A.; Neumann, S.; Ben-Hur, A.; Prenni, J. E. RAMClust: A Novel Feature Clustering Method Enables Spectral-Matching-Based Annotation for Metabolomics Data. Anal. Chem. 201486, 6812– 6817,  DOI: 10.1021/ac501530d [ACS Full Text ], [CAS], Google Scholar
  19. 19Broeckling, C.; Afsar, F.; Neumann, S.; Ben-Hur, A.; Prenni, J. RAMClustR: Mass Spectrometry Metabolomics Feature Clustering and Interpretation . R package version 1.2.2., 2021.Google Scholar(b) R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2021. accessed. http://www.R-project.org/.Google Scholar
  20. 20Broeckling, C. D.; Ganna, A.; Layer, M.; Brown, K.; Sutton, B.; Ingelsson, E.; Peers, G.; Prenni, J. E. Enabling Efficient and Confident Annotation of LC–MS Metabolomics Data through MS1 Spectrum and Time Prediction. Anal. Chem. 201688, 9226– 9234,  DOI: 10.1021/acs.analchem.6b02479 [ACS Full Text ], [CAS], Google Scholar
  21. 21Lenth, R. V. emmeans: Estimated Marginal Means, aka Least-Squares Means , R package version 1.7.0, 2021 .Google Scholar
  22. 22De Schutter, D. P.; Saison, D.; Delvaux, F.; Derdelinckx, G.; Delvaux, F. R. The Chemistry of Aging Beer. In Beer in Health and Disease Prevention; Elsevier Inc, 2008; pp 375– 388.Google Scholar
  23. 23Uchida, M.; Suga, S.; Ono, M. Improvement for Oxidative Flavor Stability of Beer-Rapid Prediction Method for Beer Flavor Stability by Electron Spin Resonance Spectroscopy. J. Am. Soc. Brew. Chem. 199654, 205– 211,  DOI: 10.1094/asbcj-54-0205 [Crossref], [CAS], Google Scholar
  24. 24Dostálek, P.; Karabín, M.; Jelínek, L. Hop Phytochemicals and Their Potential Role in Metabolic Syndrome Prevention and Therapy. Molecules 201722, 1761,  DOI: 10.3390/molecules22101761 [Crossref], [PubMed], [CAS], Google Scholar
  25. 25Bamforth, C. W. Flavour changes in beer: oxidation and other pathways. In Oxidation in foods and beverages and antioxidant applications; Elsevier Ltd, 2010, pp 424– 444. DOI: 10.1533/9780857090331.2.424 [Crossref], Google Scholar
  26. 26)(a) Robel, E. Elimination of amino acid losses with protein acid hydrolyzates due to adsorption. Anal. Biochem. 197351, 137– 145,  DOI: 10.1016/0003-2697(73)90461-2 [Crossref], [PubMed], [CAS], Google Scholar.(b) West, J. K.; Latour, R., Jr; Hench, L. L. Molecular modeling study of adsorption of poly-L-lysine onto silica glass. J. Biomed. Mater. Res. 199737, 585– 591,  DOI: 10.1002/(SICI)1097-4636(19971215)37:4<585::AID-JBM18>3.0.CO;2-7 [Crossref], [PubMed], [CAS], Google Scholar
  27. 27Croft, A. K.; Foley, M. K. Proline-rich proteins-deriving a basis for residue-based selectivity in polyphenolic binding. Org. Biomol. Chem. 20086, 1594,  DOI: 10.1039/b800365c [Crossref], [PubMed], [CAS], Google Scholar
  28. 28Yin, X. S. Malt; American Society of Brewing Chemists, 2021.Google Scholar
  29. 29You, X.; O’Keefe, S. F. Binding of volatile aroma compounds to can linings with different polymeric characteristics. Food Science & Nutrition 20186, 54– 61, Agricultural & Environmental Science Collection; Publicly Available Content Database  DOI: 10.1002/fsn3.526 [Crossref], [PubMed], [CAS], Google Scholar
  30. 30Peacock, V. E.; Deinzer, M. L. Fate of Hop Oil Components in Beer. J. Am. Soc. Brew. Chem. 198846, 104– 107,  DOI: 10.1094/ASBCJ-46-0104 [Crossref], [CAS], Google Scholar
  31. 31Wisk, T. J.; Siebert, K. J. Air Ingress in Packages Sealed with Crowns Lined with Polyvinyl Chloride. J. Am. Soc. Brew. Chem. 198745, 14– 18,  DOI: 10.1094/asbcj-45-0014 [Crossref], [CAS], Google Scholar
  32. 32Mozzon, M.; Foligni, R.; Mannozzi, C. Brewing Quality of Hop Varieties Cultivated in Central Italy Based on Multivolatile Fingerprinting and Bitter Acid Content. Foods 20209, 541,  DOI: 10.3390/foods9050541 [Crossref], [PubMed], [CAS], Google Scholar
  33. 33Steenackers, B.; De Cooman, L.; De Vos, D. Chemical transformations of characteristic hop secondary metabolites in relation to beer properties and the brewing process: A review. Food Chem. 2015172, 742– 756,  DOI: 10.1016/j.foodchem.2014.09.139 [Crossref], [PubMed], [CAS], Google Scholar

Learn More About Pasteurizing!

If you’d like to learn more about pasteurization in general, pasteurizing unique sodas, or the equipment involved, we’d love to help!

Feel free to contact us with any questions or concerns you might have, and we’ll do our best to help you. We want you to feel confident in your understanding of this vital process and how you can make it work for you!

For more information on tunnel and batch pasteurization contact PRO Engineering/Manufacturing, Inc. at[email protected]or call(414) 362-1500 and ask for Ed Michalski, CEO.   

Edward A.Michalski Bio  
Ed Michalski started his career in the beverage industry by designing stainless steel, higher flow, spray headers for Pabst Brewing. Along with the header design, he also developed a process to produce the new headers.   Ed, along with his brother David, formed PRO Engineering/Manufacturing, Inc. 

Based on what they learned by re-designing and refurbishing other manufacturers’ pasteurizers, Ed and PRO started to offer the pasteurizer marketplace superior new pasteurizers.  

PRO Engineering/Manufacturing has been designing and manufacturing tunnel pasteurizers and batch pasteurizers for over four decades.  

PRO Engineering and Manufacturing was established in 1977 by Ed and Dave Michalski as a steel fabrication shop. For more than 40 years, PRO Engineering and Manufacturing, Inc. has been delivering solutions for beverage product shelf stability and consumption safety. As leading innovators in post-fill pasteurization, our skilled team offers the best in high-quality pasteurization equipment.

Phone: 414-362-1500 | Email: [email protected]

Location: 11175 W Heather Ave, Milwaukee, WI 53224, USA