In the development of new routine methods based on vibrational spectroscopic techniques for the rapid identification of microorganisms, spectra derived from rigidly standardized protocol are used to establish a spectral database for the nonsubjective classification and identification of clinically relevant microorganisms. Thus, it is imperative that the database be comprehensive so that the natural variance of the microorganism is captured within the spectral database. A potential problem is that in recent years, it is becoming more widely accepted that microorganisms are not necessarily unicellular organisms but rather multicellular organisms able to form complex communities with specific division of tasks and population differentiation (39
). Furthermore, it is known that biofilms are elaborate structures composed of microcolonies attached to a surface and that within these microcolonies, the bacteria are organized into communities with functional heterogeneity (6
). Given that colonies of microorganisms are complex multicellular communities, it is necessary to establish at what growth stage infrared and Raman spectra should be acquired from such (micro)colonies. In so doing, any heterogeneity which can interfere with the discrimination of microorganisms can be minimized. As the aim of these new methods is to provide the clinician with laboratory results on the same day that patient material is acquired, the culture time should be kept short: for example, approximately 6 h of growth time. To gain an understanding of the spectral heterogeneity, the development of (micro)colonies was monitored over several culture times and at various positions within the (micro)colonies.
In the discussion to follow, a large part of the analysis is based on the results of subjecting the data to hierarchical clustering analysis. This is a procedure that nonsubjectively groups the input cases (i.e., the spectra) based on similarities of their properties (the spectral characteristics). When graphically displayed, the result of the analysis forms a dendrogram; the relationship between the input cases is represented by the distance at which they connect on a dissimilarity scale (e.g., Fig. A). The more similar the cases are, the smaller their connecting distance on the dissimilarity scale. Dendrograms resemble the phylogenetic trees that arise from taxonomic classification. Groups of similar members can be readily visualized. Since the spectral information reflects the biochemistry of the sample measured, the distance in the dendrograms can be interpreted as a measure of how biochemically different the various spectra are and, hence, the measurement positions within a colony.
Infrared spectra of (micro)colony imprints.
In order to gain a further understanding of the biochemical heterogeneity of (micro)colonies, vibrational spectroscopic techniques were employed. Hierarchical cluster analysis of FT-IR spectra acquired from colonies of E. coli
CIP 54.8T which were 100 μm or larger in diameter produced a dendrogram similar to that shown in Fig. A, with the spectra tending to cluster into different groups depending upon the measurement position within the colony (not shown for brevity). By examining the individual infrared spectra and calculating difference spectra, it is possible to gain a better understanding of the source of the clustering scheme observed. In Fig. A, the FT-IR spectra acquired from the center and edge of an E. coli
CIP 54.8T colony are shown. Although to the untrained eye these two spectra look remarkably similar, any differences that exist can be highlighted by taking difference spectra. The difference spectrum which results from subtracting the spectrum of the edge from that of the center is shown as well as the difference obtained from subtracting the first-derivative spectra from the two measuring positions. Since infrared bands tend to be quite broad, thereby potentially masking peak differences, the differences are more apparent in the derivative spectra. Comparison of the peak positions with those from empirical studies in the literature (24
) reveals differences in the spectral region around 1,230 cm−1
which can be assigned to the phosphate double-bond asymmetric stretching vibration of phosphodiester, free phosphate, and monoester phosphate functional groups. Smaller alterations are also observed in the protein amide I regions (approximately 1,620 to 1,670 cm−1
). This band arises predominantly from the C==O stretching vibration of the amide C==O groups of proteins. Furthermore, changes visible around 1,400 cm−1
may be attributed to the symmetric stretching vibrations of COO−
functional groups, and very weak changes were observed in the carbohydrate region around 900 to 1,200 cm−1
. Similar changes were observed for the other bacterial and yeast strains (35
), although the changes were more pronounced in the yeast strains (Fig. B).
FIG. 3 Original FT-IR spectra from a 100-μm-diameter E. coli CIP 54.8T colony (A) and a 100-μm-diameter C. albicans ATCC 90028 colony (B) measured at the center (1) and edge (2) positions of the colony. The corresponding difference spectrum (magnification, (more ...)
Interestingly, in a separate study measuring FT-IR spectra from 12-h colonies of the two E. coli strains (CIP 54.8T and 53.126), heterogeneity between spectra acquired from the center and edge of a colony was large enough to influence the discrimination between the different strains. As shown in Fig. A, the spectra arising from the two strains formed mixed clusters. However, when spectra acquired from younger colonies (approximately 7 h of culture time) were subjected to cluster analysis, two major clusters were formed corresponding to the different strains (Fig. B). These results suggest that with the older colonies, there is significant heterogeneity in the spectra from various positions within the colony. Similar studies performed with 50-μm-diameter colonies or growth time of about 6 to 7 h revealed that there was very little variance in the infrared spectra sampled from the center or periphery of the colony. Therefore, it appears that until 6 to 7 h of growth, there is very little heterogeneity observed in the composition of microcolonies. However, beyond this time frame, marked biochemical differences which vary from the center to the periphery of the colony, such as changes in the protein amide I bands, phosphate moieties likely arising from nucleic acids, protein constitution, and carbohydrate moieties, are noted. These differences likely influence the classification results observed (Fig. A and B).
Dendrogram from hierarchical clustering analysis of FT-IR spectra of 12-h cultures (A) and 7-h cultures (B) of E. coli CIP 54.8T (denoted by x) and E. coli CIP 53.126 (denoted by o). a.u., arbitrary units.
Raman spectra directly from (micro)colonies.
With infrared microspectroscopy, spectra were acquired through the entire depth of the colony at the central, intermediate, and peripheral regions. However, with this approach, any heterogeneity arising from various depths within the colony would not be readily revealed. Insight into the heterogeneity of microcolonies can also be obtained from confocal Raman microspectroscopy, in which spectra can be acquired from the various lateral positions throughout the colony as well as at various depths within the colony. The Raman spectra acquired in this manner were subjected to hierarchical cluster analysis. In Fig. A, the results are shown for spectra acquired from 6-h cultures. Visual inspection of the dendrogram revealed no obvious groupings or clusters. This observation is in accordance with the infrared findings for 6-h colonies, thereby suggesting that at this growth stage, the cultures are quite homogeneous overall. Such observations were apparent irrespective of the strain studied.
FIG. 5 Dendrograms from hierarchical cluster analysis of Raman spectra from various measurement positions within a 6-h E. coli CIP 53.126 microcolony (A) and a 24-h E. coli CIP 53.126 colony (B). Shading highlights the various clusters and corresponds to the (more ...)
Unlike the dendrograms obtained for 6-h cultures, the dendrograms of spectra acquired from microorganisms grown for 12 and 24 h showed distinct subclusters (Fig. B). When the members of the same cluster were assigned to a group and the various groups were projected onto a schematic diagram depicting the measurement location of each spectrum, it appears that there are different layers in the 12- and 24-h colonies (Fig. ). Similar findings were observed for the various strains studied for 12- and 24-h cultures. Further examination of the spectra indicates that for S. aureus
CIP 4.83, the clustering differences arise from distinct spectral peaks at 1,004, 1,158, and 1,522 cm−1
, which can be assigned to the various C—C vibrations found in carotenoids (Fig. ) (29
). The clustering reveals that the carotenoid concentration is higher within the upper layers of the colony and less prominent in the deeper layers. Carotenoids are responsible for the yellow-orange pigmentation observed in the 12- and 24-h colonies and are one of the classical characteristics of this species. Studies have shown that S. aureus
is very sensitive to the bactericidal effects of fatty acids, such as oleic acid. The incorporation of such lipophilic agents into the membranes results in increased membrane fluidity and thus in a decrease in membrane-associated functions (5
). It is believed that the production of carotenoids might help S. aureus
stabilize its cell membrane, thereby preventing potentially lethal fatty acid-induced changes in the fluidity of its membrane (5
). Other studies have also shown that pigmented S. aureus
strains are far more resistant to singlet oxygen lethality than are carotenoidless S. aureus
). Hence, the bacterium might use the carotenoid pigmentation as a mechanism to resist killing by fatty acids and to quench singlet oxygen, thus protecting against lethal effects of photosensitization. Previous studies (18
) have shown that the carotenoid production is mainly correlated with the time of growth, and this finding has also been observed by FT-IR spectroscopy (29
). Therefore, it might signify that older cells which produce significant pigmentation are found towards the surface layers of the colony. Alternatively, our finding of higher carotenoid concentration in the upper layers of older colonies might suggest a means by which the colony protects itself from its environment.
FIG. 6 Diagrammatic projection of the various clusters (, cluster 1; ○, cluster 2; , cluster 3; as determined from Fig. ) from the hierarchical cluster analysis of Raman spectra from various measurement positions within (more ...)
Averaged Raman spectra of the members of cluster 1 (surface layer) (A) and cluster 2 (layer beneath surface) (B) and the corresponding difference spectrum (A and B) from a 24-h colony of S. aureus CIP 4.83 (C) are shown. a.u., arbitrary units.
Interestingly, the Raman spectra of the other S. aureus
strain, CIP 53.154, showed that this strain does not produce the characteristic pigmentation. Nonpigmented derivatives of S. aureus
are known to exist and are often found in subcultures of stored organisms (53
). The cluster analysis shows a similar sort of distinction, with spectra acquired from the surface layers clustering together and those within deeper layers clustering as a group. However, the lack of pigmentation suggests that another spectral feature is responsible for the formation of distinct clusters. This clustering trend was found for S. aureus
CIP 53.154, E. coli
CIP 53.126, E. coli
CIP 54.8T, and C. albicans
ATCC 90028. Closer examination of the Raman difference spectra showed that in the deeper layers of 12- and 24-h colonies, there are characteristic spectral peaks at 723, 783, 813, and 1,575 cm−1
(Fig. ). These features all arise from the nucleotide and phosphate backbone vibration found in RNA (19
). It appears that the RNA concentration is higher in the deeper layers of the colony. Observations of the decrease in RNA content in older cells which have transitioned to the stationary phase from the logarithmic phase have been reported with FT-IR spectra of Bacillus subtilis
). Hence, this finding again suggests that the colony is composed of older cells in the surface layers and younger cells in the deeper layers which are more actively dividing, thus reflecting a higher RNA content.
FIG. 8 Averaged Raman spectra of the members of cluster 2 (layer beneath surface) (A) and cluster 3 (deeper layer) (B) and the corresponding difference spectrum (A and B) from a 24-h colony of E. coli CIP 54.8T (C) are shown. The Raman spectrum of RNA (D) is (more ...)
Aside from RNA differences, it was noted that 12- and 24-h colonies from the bacterial strains contain a relatively higher glycogen concentration in the surface layers (Fig. ). This glycogen difference is not observed with younger 6-h bacterial microcolonies (Fig. ) or with the yeast strain. At present it is unknown whether the glycogen is contained within the cells of the surface layers or found extracellularly in the form of a film. Previous FT-IR studies have also found increases in the carbohydrate C—O stretching mode of 24-h cultures of Bradyrhizobium japonicum
strains which have been transferred from liquid to solid culture medium. From transmission electron micrographs, the authors ascribed such changes to alteration of the bacterial wall component, possibly the formation of glycocalyx (56
). The organization of colonies into distinct layers has also been observed with E. coli
strains (cultured for 2 weeks) in which vertical sections through colonies revealed a stratification of different cell types, as could be seen with standard microscopic reagents, such as staining with toluidine blue (41
). Previous reports in the literature using scanning electron microscopy to study the surface structure of E. coli
colonies growing for over 24 h on agar medium in normal petri dishes have revealed that each colony secretes extracellular materials, some of which form a skin or framework over its surface (38
). Other studies have shown that at later stages of colony development (20 to 24 h), the surface film of E. coli
colonies became thicker. On the other hand, the film was not observed for colonies cultured for 6 to 16 h of growth (45
). Therefore, it is possible that the glycogen-rich surface layer observed with Raman microspectroscopy is the polysaccharide-rich extracellular coat, commonly known as the glycocalyx, of bacterial cells. These exopolysaccharides are mainly composed of homopolysaccharides (cellulose, levans, dextrans, and glucans) and heteropolysaccharides (monosaccharides including a uronic acid) (43
). It is thought that the formation of the glycocalyx serves as an integral matrix for a biofilm and that following the adhesion of bacteria to a substrate, the glycocalyx forms a protective milieu for cell division and microcolony formation and growth (7
). Some studies propose that the glycocalyx either acts as a diffusion barrier or, by complexing antibacterial agents, excludes and/or influences the penetration of antimicrobial agents to the underlying cells (10
). Modern medicine increasingly relies on the use of indwelling medical devices such as catheters and prosthetic joints for multiple purposes. These so-called foreign bodies are implanted for a short period, intermittently or permanently. One of the most frequently encountered complications of these devices is the development of infections. The ability of bacteria to adhere to the surface of these indwelling devices by binding to biofilm layers is still not completely understood (H. P. Endtz, personal communication). Hence, there is much interest in the development of biofilms, associated with disease in humans due to the increasing use of medical devices and the difficulty, resulting from resistance to antimicrobial agents, of effectively controlling infection (6
FIG. 9 Averaged Raman spectra of the members of cluster 1 (surface layer) (A) and cluster 2 (layer beneath surface) (B) and the corresponding difference spectrum (A and B) from a 24-h colony of E. coli CIP 53.126 (C) are shown. The Raman spectrum of glycogen (more ...)
FIG. 10 Raman spectra are shown from different depths corresponding to the measuring positions in the schematic diagram of Fig. , with A1 from the surface (A) and A3 from deeper within the colony (B) and the corresponding difference spectrum (A (more ...)
Overall, these infrared and Raman studies of the development of microorganisms cultured for various growth times reveal that there is significant colony heterogeneity in the strains cultured for 12 and 24 h. These differences can be attributed to higher glycogen content in the surface layers and to increased levels of carotenoid pigmentation in certain S. aureus
strains. Furthermore, a relatively higher RNA content was observed in the deeper layers of the colony. Therefore, spectra derived from these older colonies are quite variable, indicating the need to sample spectra from a multitude of positions within these colonies in order to capture the biological variance of the various cell types. The lack of group clusters and absence of obvious spectral differences in the various spectra obtained from 6-h cultures suggest that the microcolonies at this growth stage are very homogenous in terms of molecular composition. Thus, these spectra are suitable for inclusion in and building of spectral libraries of microorganisms. With the development of a comprehensive spectral database, it should be possible to use Raman and FT-IR microspectroscopies to provide rapid identification and classification of clinically relevant microorganisms. Moreover, the present study demonstrates that vibrational microspectroscopy can be applied to further understand the heterogeneity of microorganism growth. For example, the attachment and microcolony formation of biofilms, as well as the actual mechanisms of biofilm resistance to antimicrobial agents, still remain unclear (13
). FT-IR spectroscopy, including attenuated total reflectance spectroscopy, has been used previously to study bacterial growth and biofilm formation (56
). The use of Raman microspectroscopy to probe various layers within a colony can be extended to study the formation of sessile communities found at the base of the biofilm. These sessile cells are believed to be the root of many persistent and chronic bacterial infections since they can withstand host immune responses, unlike their nonattached planktonic counterparts which are killed by antibiotic therapy (6
). The knowledge gained from such studies can be used to develop new strategies for the treatment of infection, especially those associated with indwelling medical devices.