Single amino acid utilization for bacterial categorization

Despite great advancement in genetic typing, phenotyping is still an indispensable tool for categorization of bacteria. Certain amino acids may be essential for bacterial survival, growth, pathogenicity or toxin production, which prompts the idea that the intrinsic ability to utilize single amino acid under live-or-die situation could be a basis for differentiation of bacteria species. In this study, we determined the single amino acid consumption profiles of 7 bacterial species, and demonstrated that most bacteria have species-specific pattern of amino acid consumption. We also discovered that bacterial strains from different hosts, toxigenicity, and antibiotic-resistance presented distinct preference for certain amino acids. Taken altogether, the amino acid consumption profiles showed potential to be a novel tool complementary to study not only bacterial categorization but also biochemical characteristics of the bacteria such that its phenotyping can be used to uncover strategies for nutritional, pharmaceutical, taxonomic, and evolutionary aspects of bacterial researches.

www.nature.com/scientificreports/ strictly required for growth while Leu was not necessary but stimulated growth. The previous study of amino acid auxotrophies revealed that Arg, histidine (His), Met, tryptophan (Trp), isoleucine (Ile), Leu, and Val were essential for Enterococcus faecalis 15 . In addition, Glu and Gln were found to be an important hub in metabolism. Although these kinds of information above are not rare, they are mostly sporadic with different experimental conditions and many cultured with multiple AAs or concurrent nutritional supplements. Different bacterial species require different types of AAs for growth, development of resistance to a hostile environment, or toxin production, which means certain AAs are essential for some bacteria but may not be necessary for others. These facts pointed to a possibility to employ AA utilization as a phenotypic mechanism to distinguish bacterial species, especially further down to strains, types, toxigenicity, and other characteristic properties. However, a comprehensive study of simple or essential AA utilization among various bacterial strains is still lacking. For measuring AA utilization of bacteria, a method that can accurately quantitate AAs is needed. Flow-injection analysis (FIA) with electrochemical detection using copper nanoparticle-plated screen-printed carbon electrode (Cu n SPE) has been successfully demonstrated for sensitive determination of all native AAs 16,17 . Compared with tag-based methods 18 , this electrochemical detection method can directly detect AAs without any chromophore or fluorophore groups. Moreover, this method is more economical and easier than spectroscopic methods such as nuclear magnetic resonance (NMR) and Raman spectroscopy 19,20 . Thus, this Cu n SPE detection method can be a practical tool in the measurement of AA utilization for bacterial differentiation.
The aims of this study were to exam the hypothesis that the ability of different bacteria species to utilize single AA under minimal nutritional condition (live or die) is distinct enough to comprise a new method for novel classification/typing of bacterial species, and for study of animal tropisms in the same bacterium, differentiation of toxigenicity in the same bacterium, and differentiation of the same bacterial species with or without antimicrobial resistance. Our findings may provide pivotal phenotypic information in terms of AA dependency and may uncover strategies for nutritional, pharmaceutical, taxonomic, and evolutionary aspects of bacterial researches.

Results
Differential utilization preference of 20 AAs among bacterial species. The 7 bacterial species, namely E. coli, S. enterica, P. aeruginosa, S. aureus, S. hyicus, E. faecalis, and P. multocida, showed distinctive consumption profiles (Fig. 1). The consumption levels of AA can be conveniently categorized into high (> 80%), intermediate (20%-80%), low (< 20%), and negative (biotransformation) (< 0%). Both E. coli and S. enterica (Fig. 1a,b) have consistently high consumption of Asp, Asn, Lys, Gln, and Ser. In addition, a high consumption of Glu and a slight negative consumption of His were observed in E. coli. In most P. aeruginosa strains (Fig. 1c), Gly, Asn, Gln, Arg, Ser, and Pro were highly consumed. S. aureus and S. hyicus (Fig. 1d,e) exhibited a similar pattern in highly consuming Gly, Asn, Gln, Ser, and Thr. The E. faecalis (Fig. 1f) had an inconsistent consumption pattern in comparison with other bacterial species; only Phe and tyrosine (Tyr) were found to have a relatively steady utilization. The consumption of most AAs was low in P. multocida (Fig. 1g); only a slight negative consumption of His, Lys, and Arg was consistently observed. Overall, nearly all strains of the 7 species have low consumption of Val, Leu, Ile, and Met.

Differentiation of bacterial species by AA consumption profiles. Hierarchical clustering analysis
(HCA) on all 213 strains (Fig. 2a) revealed 7 clusters corresponded to the 7 bacterial species. All 52 strains of E. coli and all 27 strains of S. enterica were independently clustered with Glu being a significant factor to distinguish these two clusters. All P. multocida strains formed an individual cluster mainly due to the relatively low consumption of Asn, Gln, and Ser. Almost all strains of E. faecalis were grouped according to their specific consumption of Phe and Tyr. Most strains of P. aeruginosa were clustered together based on their ability to specifically consume Arg and Pro, whereas some of the strains were distributed to S. hyicus cluster. S. aureus cluster contained most of the S. aureus strains, but also contained some strains of S. hyicus and vice versa. Overall, HCA indicated that AA consumption pattern successfully categorized 100% of E. coli (52/52), S. enterica (27/27) According to the results of HCA, 11 AAs (Gly, His, Tyr, Asp, Asn, Glu, Lys, Gln, Ser, Thr, and Pro) showing higher variance among the 7 bacterial species were selected to perform principal component analysis (PCA) for further validation to differentiate/classify bacterial species (Fig. 2b). In order to confirm whether the selected principal components (PCs) were representative enough of AA consumption data, eigenvalues and the cumulative proportion of the PCs were checked ( Supplementary Fig. 1). The eigenvalues of PC1, PC2, and PC3 were all more than 1 and in combination can explain 80% of the variation, suggesting that the first 3 PCs in our PCA had an adequate reliability. The results (Fig. 2b) demonstrated that 6 species (except for S. hyicus which could not be readily distinguished from the S. aureus group) were nicely separated into 6 main groups.

Distinction by hosts.
Consumption patterns were also analyzed based on the host origin for each bacterial species. In S. aureus, disparate consumption patterns were observed between the strains from chicken and cattle (both groups were not methicillin-resistant strains). The heat map and hierarchical clustering indicated that isolates from the cattle distinctively utilize Gly, Ser, Asn, Gln, and Thr lower than those from chicken with statistically significant difference (P < 0.05) (Fig. 3)   www.nature.com/scientificreports/ Decision tree strategy based on specific AA utilization for bacterial differentiation. Based on the overall analysis of AA utilization, we discovered that 8 AAs, namely Asn, Lys, Tyr, Glu, Thr, Ser, Pro, and Gln can be used to make a decision tree for quick differentiation of these 7 bacteria species in a blind test situation (Fig. 6). These 7 bacterial species could be successfully differentiated in 4 serial decision making processes by using 50% or 75% as the cut-off consumption level (Table 1 and Fig. 6). For instance, an unknown bacterial strain is first tested by Asn, if the consumption is over 50%, it is further tested by Lys; and if the consumption is over 75%, then it can further be tested for Glu, if the Glu consumption is again over 75%, then this test bacterium is determined to be E. coli, otherwise (if the Glu consumption is less than 75%) the bacterium is deemed to be S. enterica. The decision tree can also differentiate MRSA versus chicken MSSA by the utilization level of Gln.

Discussion
In this study, an AAs consumption profile was established for phenotypic categorization of bacteria ( Fig. 1). Different from other common bacterial phenotyping methods such as API or the Biolog system, which are based on detection of enzymatic activity related to biological transformation of carbohydrates, proteins, or organic acids, www.nature.com/scientificreports/ our method directly and quantitatively analyzes the consumption levels of specific AAs. The method described here can afford a more focused fingerprinting of bacterial metabolism of AAs and complement other phenotyping methods. The ease and simplicity of this method allow rapid and continuous detection of AAs without any derivatization or complex sample processing. For reliable detection, we injected 1.5 mM Gly as an internal standard before and after the detection of each AA ( Supplementary Fig. 3), and set the error range of the internal standard within 15%. With this precision of quantitative analysis, the exact difference in consumption levels and thus their phenotypic properties, can be explicitly distinguished among bacterial species. However, the major drawback of our method is that the bacteria must survive and grow in liquid media (broth cultures); for those that can only grow on agar plates, gas phase analysis 21,22 might be an alternative. By determining each consumption level for the 20 AAs, the utilization phenotype and nutritional requirements of bacteria can be understood. Among 7 bacterial species, 5 species, namely E. coli, S. enterica, P. aeruginosa, S. aureus, and S. hyicus, showed consistently high consumption of 3 neutral polar AAs (Ser, Asn, and Gln). All 3 AAs play a crucial role in bacterial metabolism; for instance, Ser can be catabolized by L-serine dehydratase to produce pyruvate and ammonia, which is important for bacteria to regulate nitrogen balance and maintain extracellular pH 23,24 . Likewise, Gln and Asn are required as nitrogen donors for many important biosynthetic reactions in bacterial cells, including nucleic acid synthesis and transamidation reactions 25,26 . Previous works also demonstrated that Ser, Asn, and Gln could be rapidly utilized and metabolized by E. coli and other small intestinal bacteria 27,28 . All 7 species in our study showed low consumption of branched-chain amino acids (BCAAs) (Val, Leu, and Ile). Although BCAAs have been reported to play an important role in bacterial protein synthesis and maintenance of central metabolism [29][30][31] , studies in pig small intestine revealed that BCAAs are largely degraded by intestinal mucosal cells instead of the microbiota, which may suggest that either the bacteria could not utilize a great amount of BCAAs or a small amount of BCAAs would be sufficient to support the growth of these bacteria 32,33 . Thus, significant consumption of extracellular BCAAs could not be observed in our current study. Both Glu and Asp were largely consumed by E. coli in the current study; this may be related to the presence of glutamate transporters importing extracellular Glu and Asp into the cells 34 . Several L-glutamate transport  www.nature.com/scientificreports/ systems have been identified in E. coli, such as a sodium-dependent transport system for Glu, a binding proteindependent Glu/Asp transport system, and a proton symport system for Glu and Asp [35][36][37] . High levels of Lys were consumed only by E. coli and S. enterica; the possible reason might be that lysine decarboxylase's activity is enhanced for tolerating the acidic environment caused by the metabolic byproducts produced from cultivating E. coli and S. enterica 38,39 . The observation that Gly was largely consumed by S. aureus and S. hyicus might be related to the formation of pentaglycine cross bridge, which is required for linking glycan chains into a two-or three-dimensional network in the biosynthesis of the peptidoglycan cell wall, and the absence of a pentaglycine cross bridge would inhibit the survival of S. aureus cells [40][41][42] . Collectively, the preferences for AAs are consistent enough such that categorization of bacteria by their intrinsic ability to utilize a simple form of nutrient becomes a feasible new way to complement genetic typing. Further validation by HCA and PCA pointed out the critical AAs with prominent effect on bacterial differentiation (Fig. 2). Most clusters of bacterial species identified by both analyses were well-separated except for the S. hyicus and S. aureus. Both bacteria belong to the genus Staphylococcus with high genetic similarity and biochemical properties 43 . In fact, detection of S. hyicus by API 20 has been reported to show low specificity 44 , thus, it is not totally unique that they shared similar AA consumption patterns. Nevertheless, a greater consumption of Arg and Pro by most S. hyicus strains in comparison to that by S. aureus strains was observed (Fig. 1d,e), which may carry biochemical significance associated with their coagulase productivity. S. aureus and S. hyicus are classified as coagulase-positive and coagulase-variable staphylococci, respectively, based on their ability to produce extracellular coagulase that can convert fibrinogen to fibrin 45 . Studies have suggested that Arg is an important alternative substrate to coagulase-negative staphylococci 46 , hence, consumption of Arg by most S. hyicus strains may possibly be one factor contributing to the variance in coagulase production. While this speculation requires further elucidation, our AA consumption profiles showed great potential for studying the biochemical characteristics of bacteria. Some of the AA utilization determined in our study were not in agreement with those in previous reports. For instance, Met and Glu were regarded as necessary elements for P. multocida growth 14 , however, we did not observe the consumption of these AAs in our study. Similarly, a large amount of Arg was not consumed by S. aureus 10 . The discrepancy might be due to several reasons. First, the nutritional environment of the culture media may influence the metabolic state of the bacterial cells, resulting in different bacterial phenotypes 47 . In this study, the minimal nutritional medium (MNM) used for determining AA consumption only contained very minimal amounts of glucose and salt for energy source and for maintaining osmotic pressure, respectively. The nutritional sources in MNM are relatively poor in relation to media used in other studies, which possibly forced the change in bacterial metabolism. Second, the AA requirements are different during each growth phase: bacteria in the exponential phase need more AAs for cell growth and division than those in the stationary phase. Third, specific substances are required when bacteria utilize AAs,for example, S. aureus S-6 needs biotin as a growth factor when Glu is used as carbon source 48 . One interesting observation here was that some AAs presented negative consumption, which technically was due to a larger electrochemical current value of the samples in comparison to the positive control ( Supplementary Fig. 3). Taking E. coli for example, a slight negative consumption of His was consistently observed in each strain. This phenomenon may suggest that certain AA metabolite(s)  . − indicates the average consumption is less than 50%; ○ indicates the average consumption is from 50 to 75%; + indicates the average consumption is more than 75%.

Asn Glu Lys Gln Met Arg Ser Thr Cys Pro
E. coli 1-52 www.nature.com/scientificreports/ which could also produce electrochemical signals were formed upon utilization of His; further identification of metabolites is needed to confirm this deduction. It is noteworthy that the negative value was not merely a measurement of consumption but likely the sum of processes involving both the consumption and production of electroactive metabolite(s), which taken together might be better represented by the term "biotransformation. " These observations imply that the nutritional requirement of a bacterium is the result of complex evolutionary processes that cannot be deciphered or is beyond the understanding of current knowledge. Therefore, this study provided a scenario that indicates how different bacterial species react to its survival in a nutrient-scarce situation. In fact, a partial parallel study using only a single AA without MNM also showed very similar results ( Supplementary Fig. 4).
Other potential applications of the current method were noticed during the study. First, S. aureus showed distinctive consumption profiles between different hosts, i.e., cattle and chicken (Fig. 3). Almost no AAs were consumed to a significant degree by S. aureus isolated from mastitis cow milk. Although a previous study suggested that strains from milk required several AAs for growth, milk is normally deficient in free AAs 49 so alternative sources of nitrogen might be available for sustained growth of these isolates. Next, a relatively higher level of BCAA consumption was observed in MRSA and colistin-resistant E. coli compared to that in non-resistant strains (Fig. 4). Antibiotic-resistant strains were suggested to have reduced metabolism due to defective central metabolic pathways of glucose or AA (e.g. Ala) 50,51 . It is feasible that BCAAs, which also have a significant role in maintaining central metabolism, showed higher utilization by antibiotic-resistant strains. Interestingly, we found that human S. aureus and chicken MRSA strains had similar specific AAs consumption ( Fig. 1d; whether the similarity of AA consumption among hosts represents a higher possibility of disease spreading (cross-host movement remains to be elucidated, and is certainly worth further study. Finally, most of the E.coli isolates in this study consumed a high level of Glu and a prominent difference in the consumption of His and Glu was found between the Shiga toxin-producing and wild-type E. coli (Fig. 5). E. coli mutant strains have been reported to take up Glu at a much faster rate than wild-type strains, which can be attributed to changes in cell permeability 52 . The glutamate decarboxylase (GAD) system was also reported to play an important role in the resistance to multiple environmental stresses 53 . Collectively, through these preliminary observations, we believe that the AA consumption profile holds reasonably good potential as a tool, complementary to other bacterial typing methods, for studying specific bacterial characteristics/traits. A larger-scale investigation is warranted to strengthen the power of the above claims.
Based on the consumption profiles, we were able to create a simple and straightforward strategy for differentiation of bacterial species and/or strains with just a few AAs (Fig. 6). The prerequisite is to decide on the adequate consumption level for differentiation: for instance, 50% or 70% of consumption as the cut-off points. Although the efficacy could be improved by larger sample numbers, this prospective strategy should benefit bacterial identification and research on bacterial metabolism.
In conclusion, we applied FIA-ED to quantitatively determine single AA consumption profiles of bacteria and revealed their characteristic preferences in consuming AAs. AA phenotyping might be a potential way to differentiate bacterial species and with larger-scale investigations, might also reveal a feasible way to study/categorize the development of host tropisms, toxin production, and drug resistance. Additionally, this AA phenotyping determined not only the nutritional requirements but also in part the metabolism of AAs by bacteria (such as in the negative consumption). Therefore, bacterial strains may be categorized as "specific AAs utilizing strains" or "non-utilizable strains, " which places emphasis on both the nutritional and/or metabolic phenotyping regardless of their original species. Once the relationships between AA utilization and its characteristic roles in pathogenesis, toxogenesis, or resistance can be established, this categorization will be a more powerful and informative way to describe a bacterium. Expanding on the current study will offer more phenotypic information, which can afford researchers a new perspective in analyzing microbiological data derived from biochemical and molecular biological techniques. Specific relationships acquired from this kind of study could shed more light on bacterial metabolism, bacterial adaptation to host, and development of toxigenicity and antibiotic-resistance. They may also help develop novel strategies for treating infections. For example, offering specific AA(s) that are highly or differentially utilized by the resistant strains may alter their bacterial metabolism and thereby their antibiotic susceptibility 51 . Furthermore, it might be possible to develop media formulation to selectively inhibit or enhance the growth of specific bacterial strains for targeted studies. Preparation of amino acids (AAs). All AAs except for Gly were purchased from Applichem GmbH (Darmstadt, Germany). Gly was purchased from GERBU Biotechnik GmbH (Heidelberg, Germany). All AAs were dissolved in MNM to a final concentration of 3 mM (Tyr was first dissolved at 55 °C). MNM was prepared by adding 0.243 g K 2 HPO 4 , 0.234 g C 6  electrochemical analysis of amino acids. Flow-injection analysis (FIA) with electrochemical detection (FIA-ED) was used to conduct chronoamperometric experiments (i-t) and quantification of AA as previously described 17 . In brief, the FIA system consists of a Hitachi L-6200 intelligent pump drive, a Shimadzu SIL-10A Auto Injector with an interconnecting Teflon tube and a specially designed electrochemical flow cell (Zensor R&D, Taichung, Taiwan) incorporating a three-electrode system consisting of a working electrode (Cu n -SPE, copper nanoparticle-plated screen-printed electrode, geometric area of 0.2 cm@@@2), an Ag/AgCl reference electrode, and a platinum auxiliary electrode (geometric area of 0.07 cm@@@2). The detection condition using FIA system were: mobile phase: 0.01 M, pH 7 phosphate buffered saline (PBS), analytical potential: 0.15 V vs. Ag/AgCl, scan rate: 0.05 V/s, flow rate 0.5 mL/min, injection volume 20 μL. The internal standard is 1.5 mM Gly. Chronoamperometric signals for quantification were analyzed by CHI1021C electrochemical analyzer with CHI system software (CH Instruments, Austin, TX, USA). For determination of AA utilization, the percentage of AA consumption (A%) were determined by the equation

Methods
where N is the electrochemical current value of the negative control, in which bacterial strain was grown in the medium containing only MNM; P is the current value of the positive control, which contained 3 mM of each AA solution without bacterial inoculation; T is the average current value of replicate samples containing both bacteria and AA solution.
Statistical analysis. Genesis software version 1.8.1 was used to create heat maps and HCA 54 . In HCA, each AA consumption data was first standardized by z-score normalization and then used to perform average linkage distance hierarchical clustering. PCA was performed using the prcomp function in R version 3.5.3 to investigate similarities between individual samples based on correlation 55 . The rgl package was used to create interactive three-dimensional visualization of data. To assess the difference between groups, SPSS software (v.25) was used to perform the Mann-Whitney U test for non-normally distributed data. The data is represented as the median with interquartile range (IQR), and box plots were created for visualization of the data using R.