Comparative Proteomics of Meat Spoilage Bacteria Predicts Drivers for Their Coexistence on Modified Atmosphere Packaged Meat

Besides intrinsic and extrinsic factors such as antagonism for organic substrates or temperature, the storage atmosphere of meat has a high influence on the development of its initial microbiota. Specific modified atmospheres (MAs) selectively suppress growth of aerobic and anaerobic bacteria, thus reshaping the initial microbiota. As some microorganisms are more tolerant to MA, they overgrow competitors and produce metabolites that cause rejection of the product. In order to elucidate responses to different MA by means of metabolic adaptation and competition for organic substrates on meat, the typical representative meat spoilage bacteria Brochothrix (B.) thermosphacta TMW2.2101 and four lactic acid bacteria Carnobacterium (C.) divergens TMW2.1577, C. maltaromaticum TMW2.1581, Leuconostoc (L.) gelidum subsp. gelidum TMW2.1618 and L. gelidum subsp. gasicomitatum TMW2.1619 were chosen. Bacteria were grown in sterile glass bottles filled with a meat simulation medium, which was aerated constantly with either air, 100%_N2, 30%_CO2/70%_O2 or 30%_CO2/70%_N2. Growth of bacteria during incubation at 25°C and stirring at 120 rpm was monitored over 48 h and a label-free quantitative mass spectrometric approach was employed to determine changes within the bacterial proteomes in response to oxygen and carbon dioxide. Both Leuconostoc subsp. were intrinsically tolerant to MA, exhibiting no proteomic regulation of enzymes, whereas the other species provide a set of metabolic adaptation mechanism, enabling higher resistance to the detrimental effects of MA. Those mechanisms comprise: enhanced oxidative stress reduction, adjustment of the pyruvate metabolism and catabolic oxygen consumption in response to oxygen and intracellular pH homeostasis, maintenance of osmotic balance and alteration of the fatty acid composition in response to carbon dioxide. We further evaluated the potential of industrial used MA to inhibit specific bacterial spoilage. No bacterial inhibition is predicted for 30%_CO2/70%_O2 for the analyzed species, whereas 30%_CO2/70%_N2 predictively inhibits C. divergens TMW21577 and B. thermosphacta TMW2.2101. Furthermore, species-specific metabolic pathways enabling different and preferential carbon source utilization were identified, which enable non-competitive coexistence of respective bacteria on meat, resulting in synergistic spoilage. In conclusion, this study gives mechanistically explanations of their acknowledged status as typical spoilage organisms on MAP meats.

Besides intrinsic and extrinsic factors such as antagonism for organic substrates or temperature, the storage atmosphere of meat has a high influence on the development of its initial microbiota. Specific modified atmospheres (MAs) selectively suppress growth of aerobic and anaerobic bacteria, thus reshaping the initial microbiota. As some microorganisms are more tolerant to MA, they overgrow competitors and produce metabolites that cause rejection of the product. In order to elucidate responses to different MA by means of metabolic adaptation and competition for organic substrates on meat, the typical representative meat spoilage bacteria Brochothrix (B.) thermosphacta TMW2.2101 and four lactic acid bacteria Carnobacterium (C.) divergens TMW2.1577, C. maltaromaticum TMW2.1581, Leuconostoc (L.) gelidum subsp. gelidum TMW2.1618 and L. gelidum subsp. gasicomitatum TMW2.1619 were chosen. Bacteria were grown in sterile glass bottles filled with a meat simulation medium, which was aerated constantly with either air, 100%_N 2 , 30%_CO 2 /70%_O 2 or 30%_CO 2 /70%_N 2 . Growth of bacteria during incubation at 25 • C and stirring at 120 rpm was monitored over 48 h and a label-free quantitative mass spectrometric approach was employed to determine changes within the bacterial proteomes in response to oxygen and carbon dioxide. Both Leuconostoc subsp. were intrinsically tolerant to MA, exhibiting no proteomic regulation of enzymes, whereas the other species provide a set of metabolic adaptation mechanism, enabling higher resistance to the detrimental effects of MA. Those mechanisms comprise: enhanced oxidative stress reduction, adjustment of the pyruvate metabolism and catabolic oxygen consumption in response to oxygen and intracellular pH homeostasis, maintenance of osmotic balance and alteration of the fatty acid composition in response to carbon dioxide. We further evaluated the potential of industrial used MA to inhibit specific bacterial spoilage. No bacterial inhibition is predicted for 30%_CO 2 /70%_O 2 for the analyzed species, whereas

INTRODUCTION
Food packaging under different modified atmospheres (MAs) has become a common method to counteract deteriorative effects of long time storage of meat, e.g., discoloration and formation of off-odors (Yam et al., 2005;McMillin, 2008). These effects are due to growth of spoilage microorganisms, found as initial contaminations on meat. The initial microbiota is strongly influenced by distribution pathways, processing, slaughtering and storage conditions of the meat (Nychas et al., 2007(Nychas et al., , 2008. MA selectively suppress members of this initial spoilage microbiota, reshaping it to less sensitive and more tolerant species regarding MA. Those bacteria grow concomitantly upon MAP meat and induce spoilage. They are called ephemeral spoilage organisms (ESOs) as described in previous studies of Nychas et al. (2008) and comprise lactic acid bacteria (LAB), Brochothrix (B.) thermosphacta, Pseudomonas species, Enterobacterales and Shewanella and Aeromonas species (Lambert et al., 1991;Borch et al., 1996;Ercolini et al., 2006;Nychas et al., 2007;Höll et al., 2016;Hilgarth et al., 2018). As their growth rate and metabolism differs depending on the MA, shelf life can be extended and the sensorial changes can be different (Kakouri and Nychas, 1994;Esmer et al., 2011;Degirmencioglu et al., 2012).
The protective gas carbon dioxide has been described as effective against aerobic Gram-negative bacteria, e.g., Pseudomonas species (Farber, 1991;Devlieghere and Debevere, 2000). Nevertheless, retarding bacterial growth due to carbon dioxide was also described for Gram-positive bacteria such as B. thermosphacta (Molin, 1983;Devlieghere and Debevere, 2000). Despite many investigations on the bacteriostatic action of carbon dioxide, less is known yet about the molecular response to and influence of carbon dioxide on microbial metabolism (Casaburi et al., 2015). Mechanisms of action have been proposed to be an exclusion of oxygen by replacement with carbon dioxide, lowering of intracellular pH by dissociation of formed carbonic acid, an alteration of the structure of the cell membrane and an induction of osmotic unbalance (Sears and Eisenberg, 1961;Daniels et al., 1985). Furthermore, a regulatory effect of single metabolic enzymes has been demonstrated for CO 2 /HCO 3 , e.g., the intracellular level of CO 2 /HCO 3 induces virulence and toxin production in pathogens such as Helicobacter pylori and Citrobacter rodentium (Yang et al., 2009;Park et al., 2011).
However, bacterial metabolism is complex, and unknown adaptation mechanisms are hard to uncover and can be overlooked using conventional targeted methods. In contrast, novel "omic"-technologies comprising full genome, transcriptome, metabolome, or proteome analysis enable to capture a non-targeted global "snapshot" of bacterial metabolism. These techniques were recently employed to uncover metabolism, adaptation and interaction of meat spoilage bacteria (Orihuel et al., 2018;Quintieri et al., 2018;Wang et al., 2018;Höll et al., 2019). On this basis, we used a comparative full proteomic analysis, which enabled us to explore global molecular regulation mechanisms of meat spoilage bacteria toward exposure to oxygen and carbon dioxide as well as utilization of distinct organic meat-derived substrates in response to different MA.

Bacterial Strains
Representative strains for five species isolated from MA packed meat were selected among isolates of previous studies along their abundancy. B. thermosphacta TMW 2.2101 was isolated from minced beef , Carnobacterium divergens TMW 2.1577 and C. maltaromaticum TMW 2.1581 were isolated from skinless chicken breast (Höll et al., 2016) and L. gelidum subsp. gelidum TMW 2.1618 and L. gelidum subsp. gasicomitatum TMW 2.1619 were isolated from beef steaks (Hilgarth et al., 2018).

Preparation of Precultures
In order to ensure reproducibility, media of all experiments were inoculated from the same respective preculture from glycerol stock cultures prepared in brain heart infusion (BHI) media (Roth, Karlsruhe, Germany). Therefore, bacteria were grown at 25 • C aerobically in Erlenmeyer flasks or anaerobically in gas tight Schott bottles. Cultures were harvested, aliquoted, supplemented with 90% glycerol (Gerbu Biotechnik GmbH, Heidelberg, Germany) and frozen at −80 • C. Aerobic precultures were used for inoculation of experiments with air and 30%_CO 2 /70%_O 2 and anaerobic precultures were used for experiments with 100%_N 2 and 30%_CO 2 /70%_N 2 .

Experimental Setup
Bacterial cultivation was performed in gas tight locked glass bottles. Glass bottles were filled with 0.4 L MSM media, which was previously employed and described in detail by Kolbeck et al. (2019b). It is specifically adapted to components of real meat mainly consisting of meat extract, glycerol, tween80, and heme. MSM medium was inoculated with an optical density of 0.1 at 590 nm with the previous prepared precultures and stirred at 120 rpm over 48 h. Cultivation temperature was 25 ± 2 • C. During cultivation, bottles were constantly aerated with one of the four gas mixtures (air, N 2 , 30%_CO 2 /70%_O 2 , 30%_CO 2 /70%_N 2 ). Growth was monitored over 48 h by optical density measurement. Samples for proteomic analysis were taken in exponential growth phase with log CFU ml −1 > 7 to comparable to cell counts with those occurring in the relevant time frame during meat spoilage. Bacteria were cultivated in triplicates in separate glass bottles for each gas atmosphere. For proteomic data analysis, samples were taken of each replicate.

Statistical Data Analysis of the Growth
The values for lag-Phase, maximal optical density (OD max ) and maximal growth rate (µ max ) where calculated for each replicate using the open source software RStudio ver. 3.3.0 (RStudio, Inc., Boston, MA, United States) and the CRAN package grofit ver. 1.1.1-1 run with default settings. Significant differences in lag-phases, maximal optical density (OD max ) and maximal growth rate (µ max ) were analyzed between the three replicates and the gas atmospheres air, N 2 , 30%_CO 2 /70%_O 2 , 30%_CO 2 /70%_N 2 for all species by performing a one way analysis of variance (ANOVA), followed by a post hoc Tukey test assigning significant differences between means with a confidence interval of 95% (p < 0.05).

Proteomic Sample Preparation
Sample preparation was performed as described in detail previously (Kolbeck et al., 2019a). Briefly, bacterial cells were resuspended in 8M urea lysis buffer and lysed by beads beating. Total protein concentrations were determined using the BCA method. 100 µg of protein extract were reduced with 10 mM DTT and carbamidomethylated with 55 mM chloroacetamide. Subsequently, proteins were digested with 1 µg trypsin overnight at 37 • C. Digested peptide samples were desalted and resuspended in 2% acetonitrile, 98% H 2 O, 0.1% formic acid to a final concentration of 0.1 µg/µl.

LC-MS/MS Data Acquisition
LC-MS/MS measurements were performed on an Ultimate 3000 RSLCnano system coupled to a Q-Exactive HF-X mass spectrometer (Thermo Fisher Scientific). For full proteome analyses 0.5 µg of peptides were delivered to a trap column (ReproSil-pur C18-AQ, 5 µm, Dr. Maisch, 20 mm × 75 µm, self-packed) at a flow rate of 5 µL/min in HPLC grade water with 0.1% formic acid. After 10 min of loading, peptides were transferred to an analytical column (ReproSil Gold C18-AQ, 3 µm, Dr. Maisch, 450 mm × 75 µm, self-packed) and separated using a 50 min gradient from 4 to 32% of solvent B [0.1% formic acid in acetonitrile and 5% (v/v) DMSO) at 300 nL/min flow rate]. Both nanoLC solvents [solvent A = 0.1% formic acid in HPLC grade water and 5% (v/v) DMSO] contained 5% DMSO to boost MS intensity. The Q-Exactive HF-X mass spectrometer was operated in data dependent acquisition (DDA) and positive ionization mode. MS1 spectra (360-1300 m/z) were recorded at a resolution of 60,000 using an automatic gain control (AGC) target value of 3e6 and maximum injection time (maxIT) of 45 ms. Up to 18 peptide precursors were selected for fragmentation in case of the full proteome analyses. Only precursors with charge state 2 to 6 were selected and dynamic exclusion of 25 sec was enabled. Peptide fragmentation was performed using higher energy collision induced dissociation (HCD) and a normalized collision energy (NCE) of 26%. The precursor isolation window width was set to 1.3 m/z. MS2 Resolution was 15.000 with an AGC target value of 1e5 and maximum injection time (maxIT) of 25 ms (full proteome).

Statistical Data Analysis by Perseus
Either labeled free quantification (LFQ) values were used from MaxQuant for identifying significantly different regulated proteins between two sampling conditions or intensity based absolute quantification (iBAQ) values to compare general expression of proteins within each sample. LFQ and iBAQ values were identically possessed using the open source software Perseus (Tyanova et al., 2016b). Statistical data analyzation was done applying the following pipeline: (I) Data clean-up by removing proteins from potential contaminants, only identified by site or reverse, (II) log 2 data transformation and normalization, (III) removing proteins only identified in one of three replicates, (IV) calculating the mean of three replicates, (V) performing a Welch t-test of two sampling conditions (only with LFQ values), (VI) exporting data to excel (Supplementary Table S1). Afterward, proteins were filtered based on the following parameters: q-value < 0.05 and log 2 fold change > = 2. The total number of proteins significantly differentially (q-value < 0.05 and log 2 fold change > 2) up or down regulated for each species and each comparison are to be seen in Supplementary Table S2. and purine/pyrimidine and ribose metabolism Figure 6). Assignment of proteins to a specific metabolic pathway was done using functional annotations provided by the databases NCBI (Localization), RAST (Category, Subcategory, Subsystem, Role), TIGR (ROLEmain, ROLEsub1, seqdesc) and KEGG (GOnumber). Additionally, manual curation employing additional BLAST search was performed to ensure correct annotation and assignment of proteins to metabolic pathways.

Data Interpretation
Five different approaches were performed on proteins shown in Figures 1-6, to identify the effects of oxygen (21%) (Approach 1), high oxygen (70%) (Approach 2) and carbon dioxide under oxic (Approaches 3 and 5) and anoxic conditions (Approaches 4 and 5), on the proteome of bacteria ( Table 1). The effect of oxygen (21%) can be seen by focusing on the three comparisons air_vs_N 2 , air_vs_CO 2 /N 2 and CO 2 /O 2 _vs_CO 2 /N 2 (Approach 1). The effect of high oxygen concentrations (70%) can be revealed looking at the second-tier comparison of proteins differentially regulated between the two comparisons air_vs_N 2 and CO 2 /O 2 _vs_CO 2 /N 2 as well as looking at the comparison air_vs_CO 2 /O 2 (Approach 2). The effect of carbon dioxide under oxic conditions can be seen comparing the expression of proteins detected under air_vs_CO 2 /O 2 atmosphere (Approach 3). The effect of carbon dioxide under anoxic conditions can be seen from the comparison N 2 _vs_CO 2 /N 2 (Approach 4). Furthermore, a second-tier comparison of proteins differentially regulated between the two comparisons air_vs_N 2 and CO 2 /O 2 _vs_CO 2 /N 2 , also revealed the effect of carbon dioxide in response to absence or presence of oxygen (Synergistic effect, Approach 5).

Growth of Meat Spoilage Bacteria Under Different Protective Gas Atmospheres
Growth of bacteria was detected for all species under different protective gas atmospheres. Table 2 summarizes the main growth parameters OD max , lag-phase and µ max for the bacterial species. Details are provided in Supplementary Figure S1.

Metabolic Response to Carbon Dioxide Under Oxic Conditions (Approaches 3 and 5)
Various differences between species were observed concerning their metabolic response to carbon dioxide under oxic conditions. No proteomic regulation of pathways was observed for B. thermosphacta TMW2.2101, a bit more for C. divergens TMW2.1577 and both L. gelidum subspecies (up to two pathways) and a high proteomic regulation was observed for C. maltaromaticum TMW2.1581 (10 pathways). Proteins influencing intracellular osmolarity, e.g., osmotic stress proteins and ion transporters were significantly regulated for C. divergens TWM2.1577 and both L. gelidum subspecies. C. maltaromaticum TMW2.1581 exhibited an upregulation of enzymes involved in fatty acid biosynthesis and degradation as well as a downregulation of enzymes involved in sugar metabolism, adenine and ethanolamine degradation.

Metabolic Response to Carbon Dioxide Under Anoxic Conditions (Approaches 4 and 5)
Metabolic response to carbon dioxide under anoxic conditions was not detected for L. gelidum subsp. gelidum TMW2.1618. C. maltaromaticum TMW2.1581 showed again highest metabolic regulation, whereas the other three species showed little metabolic regulation to carbon dioxide. Both Carnobacteria upregulated the ADI pathway and specific enzymes for amino acid degradation in response to carbon dioxide. Additionally, C. maltaromaticum TMW2.1581 upregulated several fatty acid biosynthesis and degradation enzymes as well as enzymes for ethanolamine degradation. Latter was also upregulated by B. thermosphacta TMW2.2101 in response to carbon dioxide.

Proteomic Regulation of Respiratory Chain Enzymes
Furthermore, general non-regulated protein expression of respiratory chain enzymes was separately analyzed in response to different gas atmospheres using iBAQ values (Supplementary Table S1). All enzymes needed to establish a functional respiratory chain were expressed for all bacteria independently of the gas atmosphere.

DISCUSSION
Protective gas atmospheres comprising oxygen and carbon dioxide are commonly used to selectively inhibit microbial growth on meat (Devlieghere and Debevere, 2000;McMillin, 2008;Esmer et al., 2011). In the past, studies have aimed to reveal the effect of different gas atmospheres on the spoilage of meat, by analyzing production of metabolites, e.g., acetate, ethanol, diacetyl, 2.3-butanediol, or volatile components (Skandamis and Nychas, 2002;Balamatsia et al., 2006Balamatsia et al., , 2007Ercolini et al., 2006;Casaburi et al., 2015). Those studies focused on spoilage induced by a whole microbial consortium. Nowadays, novel methods enable a more detailed insight into the in situ FIGURE 5 | Proteomic analysis of enzymes involved in carbohydrate metabolism of five meat spoilage bacteria grown under different gas atmospheres. Data are based on three independent replicates. Proteins were classified as significantly regulated with p < 0.05 (Welch t-test). Colors indicate the effect of regulation with log 2 (diff.) = 2, log 2 (diff.) = 3, log 2 (diff.) = 4, log 2 (diff.) = 5, log 2 (diff.) = 6 and log 2 (diff.) = 7 for upregulated proteins and log 2 (diff.) = 2, log 2 (diff.) = 3, log 2 (diff.) = 4, log 2 (diff.) = 5, log 2 (diff.) = 6, log 2 (diff.) = 7, log 2 (diff.) = 8 and log 2 (diff.) = 9 for downregulated proteins. * Functional categories are based on NCBI annotation, TIGR annotation, SEED annotation, KEGG annotation, and own research. metabolism of individual meat spoilage bacteria, e.g., using a meta-transcriptomic approach as employed by Höll et al. (2019). In this study, a quantitative proteomic approach was applied to provide mechanistic insight in regulation mechanisms of individual meat spoilage bacteria to oxygen and carbon dioxide upon different gas atmosphere compositions. This approach demonstrated that individual species use different adaptation mechanisms and encode for specific metabolic pathways to cope with MAs used in meat preservation and bypass competition for nutrients by preferential use of distinct organic substrates which can be found on meat.

Effect of Different Gas Atmospheres on the Protein Expression of Respiratory Chain Enzymes
Consumption of intracellular oxygen by respiration can be considered as one major adaptation mechanism of bacteria FIGURE 6 | Proteomic analysis of enzymes involved in purine/pyrimidine and ribose metabolism of five meat spoilage bacteria grown under different gas atmospheres. Data are based on three independent replicates. Proteins were classified as significantly regulated with p < 0.05 (Welch t-test). Colors indicate the effect of regulation with log 2 (diff.) = 2, log 2 (diff.) = 3, log 2 (diff.) = 4, log 2 (diff.) = 5, log 2 (diff.) = 6 and log 2 (diff.) = 7 for upregulated proteins and log 2 (diff.) = 2, log 2 (diff.) = 3, log 2 (diff.) = 4, log 2 (diff.) = 5, log 2 (diff.) = 6, log 2 (diff.) = 7, log 2 (diff.) = 8 and log 2 (diff.) = 9 for downregulated proteins. * Functional categories are based on NCBI annotation, TIGR annotation, SEED annotation, KEGG annotation, and own research. upon high oxygen concentrations. This has been proven for the analyzed strains in a previous study that aimed to quantify oxygen consumption (Kolbeck et al., 2019b). As almost no significant differential expression of respiratory enzymes was detected in this study, we conclude that respiratory enzymes are constitutively expressed for the analyzed LABs and B. thermosphacta, enabling bacteria to immediately respond to oxygen. This is in congruent with the findings of other studies, dealing with the expression of respiratory chain enzymes of LABs (Pedersen et al., 2012).

Effect of Oxygen and Carbon Dioxide on the Metabolism of B. thermosphacta TMW 2.2101
We detected several adaptation mechanism of B. thermosphacta TMW2.2101 to the presence of oxygen, including oxygen consumption by the enzyme NADH oxidase, a downregulation of the oxygen sensitive pyruvate formate-lyase and a reduction of oxidative stress by the enzymes thioredoxin and catalase. This is in accordance to a previous study, where a high oxygen uptake rate and resistance to hydrogen peroxide was measured for the same strain of B. thermosphacta (Kolbeck et al., 2019b). Regarding an adaptation mechanism to carbon dioxide, we detected an enhanced expression of enzymes needed to degrade ethanolamine under anoxic conditions. Ethanolamine is abundant in mammalian cell membranes and can be found as part of the head group of phospholipids (Garsin, 2010). Degradation of the meat derived metabolite ethanolamine results in the production of ammonia and acetyl-CoA. As an intracellular pH-reduction is known to be caused by carbonic acid from carbon dioxide and its dissociation into carbonate and protons (Daniels et al., 1985;Smith et al., 1990), we suggest ammonia production by ethanolamine uptake and degradation as one adaptation mechanism of B. thermosphacta TMW2.2101 to carbon dioxide, which enables intracellular pH homeostasis. We further detected a prolonged lag phase for this bacterium in response to carbon dioxide exposure. As maximum OD and growth rate were not significantly different comparing air vs. 30%_CO 2 /70%_O 2 , we conclude that the inhibitory effect of carbon dioxide in 30%_CO 2 /70%_O 2 MAP meat for B. thermosphacta TMW2.2101 is foiled by the presence of high oxygen concentrations. Nevertheless, the combination of 30% carbon dioxide and absence of oxygen, needed for respiration, can effectively inhibit growth of B. thermosphacta in white meat packages as previously demonstrated by Höll et al. (2016).

Effect of Oxygen and Carbon Dioxide on the Metabolism of C. divergens TMW 2.1577
Carnobacterium divergens is one of the dominating spoilage microorganisms found on high oxygen packaged meat (Nieminen et al., 2011;Doulgeraki et al., 2012;Höll et al., 2016). As identified in this study, mechanisms of C. divergens TMW2.1577 to withstand the effect of high oxygen concentrations comprise an enhanced protein expression for reduction of oxidative stress and regulation of the pyruvate metabolism by downregulating the oxygen Analysis are based on proteomic data.
sensitive pyruvate formate-lyase. Adaptation mechanisms of C. divergens TMW2.1577 to carbon dioxide under oxic conditions comprise maintenance of the osmotic balance by upregulating several glycine betaine transporters. A disturbance of the cell membrane permeability to ionic species due to carbon dioxide has previously been described by Sears and Eisenberg (1961). Thus, by uptaking glycine betaine, present on chicken (Jung et al., 2015) and beef (Zeisel et al., 2003), bacteria are able to compensate changes in environmental osmolarity, as described by Csonka (1989). To compensate intracellular pH reduction by carbon dioxide, C. divergens TMW2.1577 increased the production of basic metabolites, e.g., ammonia and biogenic amines by upregulating the corresponding enzymes alanine dehydrogenase, tyrosine decarboxylase, and enzymes of the ADI pathway under anoxic conditions. Similar to B. thermosphacta TMW2.2101, C. divergens TMW2.1577 showed prolonged lag-phases in response to carbon dioxide but exhibited no significant difference in its total growth (OD max , µ max ) comparing air vs. 30%_CO 2 /70%_O 2 . Thus, we conclude that C. divergens TMW2.1577 is sensitive to carbon dioxide but exhibits fast proteomic adaptation prior to the lag phase and further favors high oxygen concentrations resulting in no effective inhibition by 30%_CO 2 /70%_O 2 gas atmosphere as used in red meat. Nevertheless, a significantly lower max OD was observed by comparing the gas mixture used in white meat (30%_CO 2 /70%_N 2 ) vs. air, thus assuming effective inhibition of C. divergens TMW2.1577 in white meat packages, which might not result from carbon dioxide rather than the absence of growth promoting oxygen.   4 | Summary of the effect of carbon dioxide on the lag-phase, maximal optical density OD max , maximal growth rate µ max and the proteome (− no effect, + moderate effect, ++ strong effect).

Effect of CO 2 on CO 2 Overall inhibition on
Lag-Phase µ max and OD max Proteome Inhibition and adaptation of L. gelidum subsp. gasicomitatum TMW2.1619 due to carbon dioxide. Furthermore, a significantly higher growth was detected comparing gas atmospheres used in white meat packaging (30%_CO 2 /70%_N 2 ) compared to air. Thus, we also predict no inhibition or even promoted growth of L. gelidum subsp. gasicomitatum TMW2.1619 in MAP white or red meat due to 30% CO 2 or high oxygen concentrations, respectively.

Adaptation to Different Carbon Sources of Meat-Spoiling Bacteria Under Oxic and Anoxic Conditions
There is a high diversity of different nutrients on red and white meats, which makes it unlikely that members of one species could completely occupy such a habitat outcompeting any others. This is indicated by the finding of typical consortia. Adaptation to the utilization of different carbon sources appears essential for the coexistence and microbiome dynamics of meat bacteria during product storage. In the chosen models, all bacteria utilize glucose and ribose as basic carbon sources, as could be predicted by a constitutive expression of the corresponding metabolic enzymes independent of the gas atmosphere. Beside this basic carbon sources, individual adaptation to specific other carbon sources was detected for each species. These should be even more decisive in meat spoilage as sugars are readily depleted in the initial spoilage phase employing a diverse microbiota. Regarding B. thermosphacta TMW2.2101, we detected a high lipolytic activity by utilizing myo-inositol and glycerol under oxic conditions. Furthermore, a constitutive expression of a phospholipase enzyme was measured for this species. Myoinositol and glycerol is thought to derive from the degradation of meat cells, as described by Love and Pearson (1971). Under anoxic conditions, B. thermosphacta TMW2.2101 is predicted to utilize the meat derived metabolite ethanolamine yielding energy-rich acetyl-CoA. This is in accordance to previous studies describing B. thermosphacta as major player in fatty meat products (Grau, 1983;Talon et al., 1992). Carbon sources utilized under oxic conditions by C. divergens TMW2.1577, comprise the amino acids glutamine as well as glycerol and allantoin, whereas under anoxic conditions purine and pyrimidine bases are metabolized. Enzymes for allantoin degradation could not be found in the genomes of B. thermosphacta and both Leuconostoc species. In addition to ribose, L. gelidum subsp. gasicomitatum predictively prefers pentoses as carbon source, indicated by an upregulation of several enzymes of the pentose phosphate pathway under anoxic conditions. This study contributes to a better understanding of previous research which have been done concerning microbial coexistence on meat regarding nutrition utilization and adaptation to MAs. It further evaluates the potential of the industrial used protective gases for red and white meat packaging in order to prevent meat spoilage by the species analyzed.

CONCLUSION
Protective gases such as oxygen and carbon dioxide are used in food industry to prevent bacterial cell growth and spoilage on meat. Nevertheless, we demonstrated that different spoilers encode for different metabolic pathways to cope with the detrimental effects of oxygen and carbon dioxide on their metabolism. By controlled regulation of those metabolic pathways, bacteria developed strategies to cope with high oxygen amounts, e.g., consumption of oxygen, reduction of oxidative stress or regulation of oxygen sensitive enzymes as well as high amounts of carbon dioxide, e.g., maintenance of intracellular pH, maintenance of osmotic balance and adaptation of the cell membrane by altering the fatty acid composition. Thus, the usage of MAs only prevents bacterial spoilage to a certain degree.
Furthermore, it appears that high oxygen concentrations (70%_O 2 ) antagonize with high CO 2 concentrations (30%_CO 2 ), i.e., counteracting the desired inhibitory effect of carbon dioxide on bacteria upon MAP. This effect could be demonstrated in this study for B. thermosphacta and C. divergens, which are inhibited by CO 2 only in the absence of high oxygen concentrations. No inhibition by CO 2 was demonstrated for both L. gelidum subspecies and C. maltaromaticum enabling their growth on meat independent of the MA applied. Proteomic data also demonstrated that different spoilers utilize distinct organic substrates available on meat explaining their coexistence as a microbial spoilage consortium on meat.

DATA AVAILABILITY STATEMENT
All LC-MS/MS data files and MaxQuant output files have been deposited to the ProteomeXchange Consortium (http:// proteomecentral.proteomexchange.org) via the PRIDE partner repository with the dataset identifier PXD016382.

AUTHOR CONTRIBUTIONS
SK designed the study, performed the experiments and data evaluation, and wrote the first draft of the manuscript. CL performed the mass spectrometric data acquisition. CM supervised the data evaluation. MH helped to draft the manuscript, assisted in data interpretation, and supervised the work of SK. RV initiated the project and supervised the work of SK. All authors read and approved the final manuscript.