- Email: [email protected]
A simple imaging method for biomass determination
Journal of Microbiological Methods 60 (2005) 135 – 140 www.elsevier.com/locate/jmicmeth
Note
A simple imaging method for biomass determination Carla C.C.R. de Carvalhoa,*, Marco P.C. Marquesb, Pedro Fernandesa,b,c, M. Manuela R. da Fonsecaa a
Centro de Engenharia Biolo´gica e Quı´mica, Instituto Superior Te´cnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugal b Universidade Luso´fona de Humanidades e Tecnologias, Av. Campo Grande 376, 1749-024 Lisboa, Portugal c BioTrend Lda, Rua Torcato Jorge, 41, c/v Dta, 2675-807 Ramada, Potugal Received 14 September 2004; received in revised form 27 September 2004; accepted 28 September 2004
Abstract An inexpensive and fast method based on images taken during growth of bacterial cells on multi-well plates was developed for biomass quantification. A correlation of 85% between the results obtained by image analysis and optical density measurements was obtained. This simple method allows the assessment of growth with highly aggregated cell cultures and the rapid screening of a large number of carbon sources. D 2004 Elsevier B.V. All rights reserved. Keywords: Growth quantification; Substrate screening; Image analysis; Cell clustering; Toxic substrates
There are numerous methods currently available for the determination of biomass. Among these, plate counting and direct counting procedures using microscopic methods are widely used. The differences between the results obtained with these two types of methods depend on the number of dead and/or nonviable cells, i.e., on the number of cells which are not able to form colonies, and also on both the selectivity of the growth medium and the incubation conditions used for the agar plates. In the majority of the papers published in the last two decades, fluorescent methods
* Corresponding author. Tel.: +351 21 8417681; fax: +351 21 8419062. E-mail address: [email protected] (C.C.C.R. de Carvalho). 0167-7012/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.mimet.2004.09.014
were the preferred technique due to their high sensitivity (Poglazova et al., 1996). Epifluorescence microscopy is generally acknowledged to be one of the most adequate methods for the quantification of microorganisms in all habitats (Fry, 1990; Kepner and Pratt, 1994). However, this technique has also several drawbacks, such as the inability to carry out further studies on the microorganisms used in the observations. The direct epifluorescence filter technique, in which the microorganisms are stained with dyes such as acridine orange has been accepted as a quantitative method for determining the number of bacteria in aquatic environment (Heldal et al., 1994). However, the visual counting of bacteria on filters is laborious and time-consuming. To overcome this, automated methods were developed. The most promising were
136
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
cyto- and spectrofluorimetric methods (Paul and Myers, 1982; Poglazova et al., 1984; Poglazova et al., 1996), especially flow cytometry (Tyndall et al., 1985; Donelly and Gaigent, 1986; Alcon et al., 2004; Laplace-Builhe´ et al., 1993). Quantification of specific populations is achieved by molecular techniques, such as fluorescent antibody and in situ hybridization techniques, which allow direct detection and in situ identification of microorganisms (important to assess the existence of subpopulations in biofilms). In this case, the reliability of the techniques depend on the physiological state and on the general properties of a particular strain. Besides being expensive methods, for having automated image analysis, high-quality images must be acquired with the inherent costs of the necessary imageacquisition hard and software. The ability of a bacterial strain to degrade a certain substrate, e.g., hydrocarbon or alcohol, can and has been successfully used for bioremediation and biocatalytic purposes. However, the majority of these substrates are toxic and affect the cellular membrane (de Carvalho et al., in press). This can lead to modifications at the cellular membrane level and thus in cell hydrophobicity, altering the cell adhesion properties. Often clusters of cells are formed and consequently biomass growth cannot be assessed by turbidity measurements, since the clusters absorb and scatter light differently as compared to their free cell counterparts. In the present study, we describe a simple and fast method to assess cell growth in the presence of alcohols and hydrocarbons. As shown below, cell growth could be monitored by taking photographs of the plates where cells were growing and the corresponding growth rates could be calculated by image analysis. The microorganism chosen was Rhodococcus erythropolis DCL14, which has been found able to degrade a large number of hydrocarbons and alcohols and even fuel oil. When this strain was cultured in shaken flasks, cell clustering was observed in the presence of the majority of these carbon sources (de Carvalho and da Fonseca, in press). In the present work, growth was carried out in plates with ninety-six 300-Al wells, containing 150 Al of mineral medium (Wiegant and de Bont, 1980), and in plates with twenty-four 2.75-ml wells, containing 400 Al of mineral medium. Three wells were used per
carbon source concentration. The initial carbon source concentrations tested were 0.125% and 0.25% (v/v). The organic solvents used as sole carbon and energy sources were ethanol (99.8%), butanol (N99.5%), propanol (N99.5%), n-dodecanol, cyclohexane (N99.5%) and toluene (N99.5%) from Merck; n-octane (N99%) purchased from Merck-Schuchardt; methanol (N99.8%), n-hexane (N99%) and iso-octane (N99.5%) from Riedel-de HaJn; n-undecane (99%), n-tetradecane (99%) and n-hexadecane (99%) purchased from Sigma; cyclohexanol (99%) and n-dodecane (N99%) from Aldrich; pentane (99%) purchased from Fluka; n-heptane (95%) from Lab-Scan; and n-nonane (99%) from Acros. Growth was carried out at 28 8C and 200 rpm in a Heidolph Inkubator 1000 and the optical density was measured, without further delays to avoid cell aggregation, at 600 nm with a Spectra Max 340 PC from Molecular Devices. When cell clusters were observed, cells could be separated before carrying out optical density measurements by increasing agitation to 600 rpm for 30 s. The plates were photographed seven times during the growth time course, after being allowed to settle for 30 s, using a COHU camera with a Cosmicar TV 12.5–75 mm zoom lens. The lens aperture was set to 1.8 to have as little depth-of-field as possible and the zoom was fixed at 60 mm to decrease image distortion. All images were grabbed at the same magnification in the Red–Green–Blue system. The acquisition software was Matrox Inspector 2.1. Four images of each plate were taken at each time: the corners of the plate were photograph separately and the wells that were common to more than one picture were used as controls during image analysis. Fig. 1a,b shows examples of the photographs grabbed during the experiments. The plates were agitated in a rotary shaker incubator, as previously mentioned. Due to the high hydrophobicity of R. erythropolis cells (de Carvalho et al., 2000; de Carvalho and da Fonseca, 2004), they were mainly positioned at the surface or inside the solvent hydrophobic droplets (de Carvalho and da Fonseca, 2003). These, in turn, were mainly located at the surface of the liquid phase since they have densities lower than water. Furthermore, on the surface, the access to oxygen was also easier. Thus, cell growth could be observed, under the naked eye, mainly at the surface of the aqueous phase. When
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
137
Fig. 1. Images taken during R. erythropolis growth on alcohols and hydrocarbons. (a) 96-well plate, (b) 24-well plate.
cells aggregated, the clustering occurred from the centre towards the boundary of the well (Fig. 1). Almost no cell clustering occurred in the presence of alcohols, but extended cell aggregation was observed in the presence of hydrophobic hydrocarbons such as n-dodecane, n-tetradecane and n-hexadecane (data not shown). Due to the type of agitation, the cluster formed grew mainly on the xy-plane, its thickness being almost negligible when compared to its size. Bishop and Rittmann (1995) suggested that onedimensional models might be sufficient for description of biofilms, but multidimensional modelling could be required to predict its heterogeneity. In the present case, we are only interested in the extent of cell growth and thus a simple model can be applied. Two programs were used to analyse the images acquired: Adobe Photoshop 7.0 from Adobe Systems Incorporated and Visilog 5 from Noesis SA. The former is a well known and common imaging software and was tested to assess the simplicity of the method. The procedure to analyse the images using Adobe Photoshop 7.0 was the following: the colour information of the Red–Green–Blue system was discarded and the images were transformed into grayscale pictures; brightness/contrast was adjusted to increase the differences between the cells and the image background; using the selection tool each well was selected; the colour range correspondent to the cells was selected using the colour range command and the selected pixels were counted with the histogram tool. The area of the selected well was also calculated and the results are represented in terms of barea occupied with cells per area of wellQ to avoid errors due to image focusing (which might result in slightly different camera-object distances) and different selections of the wells to be analysed between images. Imaging analysis using Visilog 5 was carried out as follows: all wells in each image were selected
and a correspondent binary image marking their position was created; the original images in the Red–Green–Blue system were transformed into the Hue–Intensity–Saturation colour system and automated segmentation was performed on the Intensity image to produce a binary image with the cells as objects; after superposition of the binary and transformed images, the pixels correspondent to cells on each well were calculated and the results were written on a text file. The results obtained with image analysis were found to correlate linearly with the turbidity measurements (Fig. 2). All data acquired with the ninety six wells plate are represented, giving a correlation of 85%. If some data were removed, the correlation would obviously increase. Nevertheless, we wanted to show that, considering our whole experimental data set, a 85% correlation could be attained. The boccupied areaQ represents the ratio between the number of pixels correspondent to cells and the number of pixels corresponding to the well in which the cells were, and is thus dimensionless. This variable can be converted to cell density using the equation given in Fig. 2, which relates O.D. with the occupied area, and a calibration curve relating the O.D. with cell density (not shown). In the case of cultures forming clusters of cells with a 3D-structure, the correlation between the optical density data and the photographic method should be checked, as part of the biomass may cover inner layers of cells. If the bacteria to test show some colour, calibration of the method may be achieved using a colourimetric scale. The highest the colour saturation, the highest the number of cells present. The image analysis data represented in Fig. 2 was acquired with Adobe Photoshop 7.0. The differences between the results collected with Adobe Photoshop and Visilog 5 were around 2% (data not shown). The
138
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
Fig. 2. Relation between the optical density (O.D.) measurements at 600 nm and the area occupied by the cells in the wells obtained by image analysis.
differences may be ascribed to variations in the manual selection of the area of the wells to be analysed and mainly to the segmentation command. In Adobe Photoshop, thresholding was carried out by the operator, whilst in Visilog entropy thresholding was performed automatically by the software. The photographic method showed a considerable sensitivity, being able to distinguish between samples with O.D. differences higher than 0.01. When two growth curves of duplicate cultures obtained with turbidity measurements (O.D.) were compared with the results attained for one of these by
fluorescence microscopy (N cells) and, by the photographic method described (occupied area), it was observed that the growth curves presented similar trends (Fig. 3). The example refers to cultures on 0.125% (v/v) ethanol. Two growth curves obtained by O.D. measurements are represented to show that cultures carried out at the same time and under similar conditions, with cells from the same inoculum, may present slightly different curves. After normalisation of the data, an average standard deviation of 0.085 was attained between the measurements made with the three methods at a certain time during the time
Fig. 3. Growth curves of R. erythropolis cells, on 0.125% (v/v) ethanol, obtained by turbidity measurements (O.D.) at 600 nm, photographic method combined with image analysis (occupied area) and fluorescence microscopy combined with image analysis (N cells, which represents the average number of cells observed per image taken). Two growth curves obtained by O.D. measurements are represented to show the variability of this method.
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
139
Fig. 4. Growth rates attained with R. erythropolis cells, in plates with 96 (96W) and 24 (24W) wells, in the presence of an initial concentration of 0.125% or 0.25% (v/v) of alcohols and hydrocarbons, as sole carbon and energy sources.
course of a culture. The major differences were observed after 40 h of cell growth due to the specificities of each method. In fact, cell death is not detected by turbidity measurements as cell debris scatter light. On the other hand, fluorescence microscopy techniques allow measurements at the individual level (Davey and Kell, 1996) and burst cells are not counted. A decrease in the average number of cells observed per image was observed from 32 h onwards due to cell death (Fig. 3). Surprisingly, the photographs of plates also revealed that cell death had occurred, which corresponded to a decrease of the area occupied by the cells in the wells (Fig. 3). This method thus gave more accurate results than the traditional turbidity measurements. The growth rates achieved with each carbon source were calculated with the data collected by the imaging method, indicating which carbon sources were better metabolised by R. erythropolis cells. Within the alcohols, the highest growth rates were achieved with ethanol and methanol (Fig. 4). With regard to hydrocarbons, n-dodecane, n-tetradecane and n-hexadecane allowed the highest growth rates, possibly due to their lower toxicity when compared to the other carbon sources tested. Nevertheless, the growth rates attained on each alcohol and hydrocarbon depended on its metabolic pathway, as seen by the low growth rates observed in the presence of hydrocarbons with an odd number of carbon atoms. In conclusion, the proposed photographic method, which consists of taking photographs of the plates where cells are grown in the presence of different
substrates, combined with relatively simple image analysis, provided growth data comparable to data obtained by turbidity measurements and by fluorescence microscopy. The method is carried out at the one-tenth millilitre scale. It thus represents a huge saving in laboratory facilities, consumables and labour, enabling a cheap and fast screening of growth substrates. In addition, the method avoids sampling and is non-intrusive. A digital camera with good resolution and an imaging software is all it requires, in contrast with the currently used methods involving expensive microscopy techniques. Furthermore, it may be used when cell clustering occurs, contrarily to optical density measurements.
Acknowledgement This study was supported by a post-doctoral grant (SFRH/BPD/14426/2003) awarded to Carla da C.C.R. de Carvalho by Fundac¸ a˜ o para a Cieˆ ncia e a Tecnologia, Portugal.
References Alcon, A., Santos, V.E., Casas, J.A., Garcı´a-Ochoa, F., 2004. Use of flow cytometry for growth structured kinetic model development. Application to Candida bombicola growth. Enzyme Microb. Technol. 34, 399 – 406. Bishop, P., Rittmann, B.E., 1995. Modelling heterogeneity in biofilms: report of the discussion session. Water Sci. Technol. 32, 263 – 265.
140
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
Davey, H.M., Kell, D.B., 1996. Flow cytometry and cell sorting of heterogeneous microbial populations: the importance of singlecell analyses. Microbiol. Rev. 60, 641 – 696. de Carvalho, C.C.C.R., da Fonseca, M.M.R., 2003. A simple method to observe organic solvent drops with a standard optical microscope. Microsc. Res. Tech. 60, 465 – 466. de Carvalho, C.C.C.R., da Fonseca, M.M.R., 2004. Principal component analysis applied to bacterial cell behaviour in the presence of organic solvents. Biocatal. Biotransform. 23, 1 – 12. de Carvalho, C.C.C.R., van Keulen, F., da Fonseca, M.M.R., 2000. Biotransformation of limonene-1,2-epoxide to limonene-1,2diol by Rhodococcus erythropolis cells: an introductory approach to selective hydrolysis and product separation. Food Technol. Biotechnol. 38, 181 – 185. de Carvalho, C.C.C.R., Parreþo-Marchante, B., Neumann, G., da Fonseca, M.M.R., Heipieper, H.J. (in press) Adaptation of Rhodococcus erythropolis DCL14 to growth on n-alkanes, alcohols and terpenes. Appl. Microbiol. Biotechnol. DOI:10.1007/s00253-004-1750-z. Donelly, C.W., Gaigent, G.J., 1986. Method for flow cytometric detection of Listeria monocytogenes in milk. Appl. Environ. Microbiol. 52, 689 – 695. Fry, J.C., 1990. Direct methods and biomass estimation. Methods Microbiol. 22, 41 – 85. Heldal, M., Norland, S., Riemann, B., Bratbak, G., 1994. Determination of bacterial cell number and cell volume by
means of flowcytometry. TEM and epifluorescence microscopy. J. Microbiol. Methods 20, 255 – 263. Kepner, R.L., Pratt, J.R., 1994. Use of fluorochromes for direct enumeration of total bacteria in environmental samples: past and present. Microbiol. Rev. 58, 603 – 615. Laplace-Builhe´, C., Hahne, K., Hunger, W., Tirilly, Y., Drocourt, J.L., 1993. Application of flow cytometry to rapid microbial analysis in food and drinks industries. Biol. Cell 78, 123 – 128. Paul, J.H., Myers, B., 1982. Fluorometric determinatiion of DNA in aquatic microorganisms by use of Hoechst 33258. Appl. Environ. Microbiol. 43, 1393 – 1399. Poglazova, M.N., Pomoshchnikova, N.A., Meisel, M.N., 1984. Determination of primary amines and protein using fluorescamine in suspensions of microorganisms. Microbiology 53, 20 – 23. Poglazova, M.N., Mitskevich, I.N., Kuzhinovsky, V.A., 1996. A spectrofluorimetric method for the determination bacterial counts in environmental samples. J. Microbiol. Methods 24, 211 – 218. Tyndall, R.L., Hand, R.E., Mann, R.C., Jernigan, R., 1985. Application of flow detection and characterization of Legionella. Environ. Microbiol. 49, 852 – 857. Wiegant, W.M., de Bont, J.A.M., 1980. A new route for ethylene glycol metabolism in Mycobacterium E44. J. Gen. Microbiol. 120, 325 – 331.
Note
A simple imaging method for biomass determination Carla C.C.R. de Carvalhoa,*, Marco P.C. Marquesb, Pedro Fernandesa,b,c, M. Manuela R. da Fonsecaa a
Centro de Engenharia Biolo´gica e Quı´mica, Instituto Superior Te´cnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugal b Universidade Luso´fona de Humanidades e Tecnologias, Av. Campo Grande 376, 1749-024 Lisboa, Portugal c BioTrend Lda, Rua Torcato Jorge, 41, c/v Dta, 2675-807 Ramada, Potugal Received 14 September 2004; received in revised form 27 September 2004; accepted 28 September 2004
Abstract An inexpensive and fast method based on images taken during growth of bacterial cells on multi-well plates was developed for biomass quantification. A correlation of 85% between the results obtained by image analysis and optical density measurements was obtained. This simple method allows the assessment of growth with highly aggregated cell cultures and the rapid screening of a large number of carbon sources. D 2004 Elsevier B.V. All rights reserved. Keywords: Growth quantification; Substrate screening; Image analysis; Cell clustering; Toxic substrates
There are numerous methods currently available for the determination of biomass. Among these, plate counting and direct counting procedures using microscopic methods are widely used. The differences between the results obtained with these two types of methods depend on the number of dead and/or nonviable cells, i.e., on the number of cells which are not able to form colonies, and also on both the selectivity of the growth medium and the incubation conditions used for the agar plates. In the majority of the papers published in the last two decades, fluorescent methods
* Corresponding author. Tel.: +351 21 8417681; fax: +351 21 8419062. E-mail address: [email protected] (C.C.C.R. de Carvalho). 0167-7012/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.mimet.2004.09.014
were the preferred technique due to their high sensitivity (Poglazova et al., 1996). Epifluorescence microscopy is generally acknowledged to be one of the most adequate methods for the quantification of microorganisms in all habitats (Fry, 1990; Kepner and Pratt, 1994). However, this technique has also several drawbacks, such as the inability to carry out further studies on the microorganisms used in the observations. The direct epifluorescence filter technique, in which the microorganisms are stained with dyes such as acridine orange has been accepted as a quantitative method for determining the number of bacteria in aquatic environment (Heldal et al., 1994). However, the visual counting of bacteria on filters is laborious and time-consuming. To overcome this, automated methods were developed. The most promising were
136
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
cyto- and spectrofluorimetric methods (Paul and Myers, 1982; Poglazova et al., 1984; Poglazova et al., 1996), especially flow cytometry (Tyndall et al., 1985; Donelly and Gaigent, 1986; Alcon et al., 2004; Laplace-Builhe´ et al., 1993). Quantification of specific populations is achieved by molecular techniques, such as fluorescent antibody and in situ hybridization techniques, which allow direct detection and in situ identification of microorganisms (important to assess the existence of subpopulations in biofilms). In this case, the reliability of the techniques depend on the physiological state and on the general properties of a particular strain. Besides being expensive methods, for having automated image analysis, high-quality images must be acquired with the inherent costs of the necessary imageacquisition hard and software. The ability of a bacterial strain to degrade a certain substrate, e.g., hydrocarbon or alcohol, can and has been successfully used for bioremediation and biocatalytic purposes. However, the majority of these substrates are toxic and affect the cellular membrane (de Carvalho et al., in press). This can lead to modifications at the cellular membrane level and thus in cell hydrophobicity, altering the cell adhesion properties. Often clusters of cells are formed and consequently biomass growth cannot be assessed by turbidity measurements, since the clusters absorb and scatter light differently as compared to their free cell counterparts. In the present study, we describe a simple and fast method to assess cell growth in the presence of alcohols and hydrocarbons. As shown below, cell growth could be monitored by taking photographs of the plates where cells were growing and the corresponding growth rates could be calculated by image analysis. The microorganism chosen was Rhodococcus erythropolis DCL14, which has been found able to degrade a large number of hydrocarbons and alcohols and even fuel oil. When this strain was cultured in shaken flasks, cell clustering was observed in the presence of the majority of these carbon sources (de Carvalho and da Fonseca, in press). In the present work, growth was carried out in plates with ninety-six 300-Al wells, containing 150 Al of mineral medium (Wiegant and de Bont, 1980), and in plates with twenty-four 2.75-ml wells, containing 400 Al of mineral medium. Three wells were used per
carbon source concentration. The initial carbon source concentrations tested were 0.125% and 0.25% (v/v). The organic solvents used as sole carbon and energy sources were ethanol (99.8%), butanol (N99.5%), propanol (N99.5%), n-dodecanol, cyclohexane (N99.5%) and toluene (N99.5%) from Merck; n-octane (N99%) purchased from Merck-Schuchardt; methanol (N99.8%), n-hexane (N99%) and iso-octane (N99.5%) from Riedel-de HaJn; n-undecane (99%), n-tetradecane (99%) and n-hexadecane (99%) purchased from Sigma; cyclohexanol (99%) and n-dodecane (N99%) from Aldrich; pentane (99%) purchased from Fluka; n-heptane (95%) from Lab-Scan; and n-nonane (99%) from Acros. Growth was carried out at 28 8C and 200 rpm in a Heidolph Inkubator 1000 and the optical density was measured, without further delays to avoid cell aggregation, at 600 nm with a Spectra Max 340 PC from Molecular Devices. When cell clusters were observed, cells could be separated before carrying out optical density measurements by increasing agitation to 600 rpm for 30 s. The plates were photographed seven times during the growth time course, after being allowed to settle for 30 s, using a COHU camera with a Cosmicar TV 12.5–75 mm zoom lens. The lens aperture was set to 1.8 to have as little depth-of-field as possible and the zoom was fixed at 60 mm to decrease image distortion. All images were grabbed at the same magnification in the Red–Green–Blue system. The acquisition software was Matrox Inspector 2.1. Four images of each plate were taken at each time: the corners of the plate were photograph separately and the wells that were common to more than one picture were used as controls during image analysis. Fig. 1a,b shows examples of the photographs grabbed during the experiments. The plates were agitated in a rotary shaker incubator, as previously mentioned. Due to the high hydrophobicity of R. erythropolis cells (de Carvalho et al., 2000; de Carvalho and da Fonseca, 2004), they were mainly positioned at the surface or inside the solvent hydrophobic droplets (de Carvalho and da Fonseca, 2003). These, in turn, were mainly located at the surface of the liquid phase since they have densities lower than water. Furthermore, on the surface, the access to oxygen was also easier. Thus, cell growth could be observed, under the naked eye, mainly at the surface of the aqueous phase. When
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
137
Fig. 1. Images taken during R. erythropolis growth on alcohols and hydrocarbons. (a) 96-well plate, (b) 24-well plate.
cells aggregated, the clustering occurred from the centre towards the boundary of the well (Fig. 1). Almost no cell clustering occurred in the presence of alcohols, but extended cell aggregation was observed in the presence of hydrophobic hydrocarbons such as n-dodecane, n-tetradecane and n-hexadecane (data not shown). Due to the type of agitation, the cluster formed grew mainly on the xy-plane, its thickness being almost negligible when compared to its size. Bishop and Rittmann (1995) suggested that onedimensional models might be sufficient for description of biofilms, but multidimensional modelling could be required to predict its heterogeneity. In the present case, we are only interested in the extent of cell growth and thus a simple model can be applied. Two programs were used to analyse the images acquired: Adobe Photoshop 7.0 from Adobe Systems Incorporated and Visilog 5 from Noesis SA. The former is a well known and common imaging software and was tested to assess the simplicity of the method. The procedure to analyse the images using Adobe Photoshop 7.0 was the following: the colour information of the Red–Green–Blue system was discarded and the images were transformed into grayscale pictures; brightness/contrast was adjusted to increase the differences between the cells and the image background; using the selection tool each well was selected; the colour range correspondent to the cells was selected using the colour range command and the selected pixels were counted with the histogram tool. The area of the selected well was also calculated and the results are represented in terms of barea occupied with cells per area of wellQ to avoid errors due to image focusing (which might result in slightly different camera-object distances) and different selections of the wells to be analysed between images. Imaging analysis using Visilog 5 was carried out as follows: all wells in each image were selected
and a correspondent binary image marking their position was created; the original images in the Red–Green–Blue system were transformed into the Hue–Intensity–Saturation colour system and automated segmentation was performed on the Intensity image to produce a binary image with the cells as objects; after superposition of the binary and transformed images, the pixels correspondent to cells on each well were calculated and the results were written on a text file. The results obtained with image analysis were found to correlate linearly with the turbidity measurements (Fig. 2). All data acquired with the ninety six wells plate are represented, giving a correlation of 85%. If some data were removed, the correlation would obviously increase. Nevertheless, we wanted to show that, considering our whole experimental data set, a 85% correlation could be attained. The boccupied areaQ represents the ratio between the number of pixels correspondent to cells and the number of pixels corresponding to the well in which the cells were, and is thus dimensionless. This variable can be converted to cell density using the equation given in Fig. 2, which relates O.D. with the occupied area, and a calibration curve relating the O.D. with cell density (not shown). In the case of cultures forming clusters of cells with a 3D-structure, the correlation between the optical density data and the photographic method should be checked, as part of the biomass may cover inner layers of cells. If the bacteria to test show some colour, calibration of the method may be achieved using a colourimetric scale. The highest the colour saturation, the highest the number of cells present. The image analysis data represented in Fig. 2 was acquired with Adobe Photoshop 7.0. The differences between the results collected with Adobe Photoshop and Visilog 5 were around 2% (data not shown). The
138
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
Fig. 2. Relation between the optical density (O.D.) measurements at 600 nm and the area occupied by the cells in the wells obtained by image analysis.
differences may be ascribed to variations in the manual selection of the area of the wells to be analysed and mainly to the segmentation command. In Adobe Photoshop, thresholding was carried out by the operator, whilst in Visilog entropy thresholding was performed automatically by the software. The photographic method showed a considerable sensitivity, being able to distinguish between samples with O.D. differences higher than 0.01. When two growth curves of duplicate cultures obtained with turbidity measurements (O.D.) were compared with the results attained for one of these by
fluorescence microscopy (N cells) and, by the photographic method described (occupied area), it was observed that the growth curves presented similar trends (Fig. 3). The example refers to cultures on 0.125% (v/v) ethanol. Two growth curves obtained by O.D. measurements are represented to show that cultures carried out at the same time and under similar conditions, with cells from the same inoculum, may present slightly different curves. After normalisation of the data, an average standard deviation of 0.085 was attained between the measurements made with the three methods at a certain time during the time
Fig. 3. Growth curves of R. erythropolis cells, on 0.125% (v/v) ethanol, obtained by turbidity measurements (O.D.) at 600 nm, photographic method combined with image analysis (occupied area) and fluorescence microscopy combined with image analysis (N cells, which represents the average number of cells observed per image taken). Two growth curves obtained by O.D. measurements are represented to show the variability of this method.
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
139
Fig. 4. Growth rates attained with R. erythropolis cells, in plates with 96 (96W) and 24 (24W) wells, in the presence of an initial concentration of 0.125% or 0.25% (v/v) of alcohols and hydrocarbons, as sole carbon and energy sources.
course of a culture. The major differences were observed after 40 h of cell growth due to the specificities of each method. In fact, cell death is not detected by turbidity measurements as cell debris scatter light. On the other hand, fluorescence microscopy techniques allow measurements at the individual level (Davey and Kell, 1996) and burst cells are not counted. A decrease in the average number of cells observed per image was observed from 32 h onwards due to cell death (Fig. 3). Surprisingly, the photographs of plates also revealed that cell death had occurred, which corresponded to a decrease of the area occupied by the cells in the wells (Fig. 3). This method thus gave more accurate results than the traditional turbidity measurements. The growth rates achieved with each carbon source were calculated with the data collected by the imaging method, indicating which carbon sources were better metabolised by R. erythropolis cells. Within the alcohols, the highest growth rates were achieved with ethanol and methanol (Fig. 4). With regard to hydrocarbons, n-dodecane, n-tetradecane and n-hexadecane allowed the highest growth rates, possibly due to their lower toxicity when compared to the other carbon sources tested. Nevertheless, the growth rates attained on each alcohol and hydrocarbon depended on its metabolic pathway, as seen by the low growth rates observed in the presence of hydrocarbons with an odd number of carbon atoms. In conclusion, the proposed photographic method, which consists of taking photographs of the plates where cells are grown in the presence of different
substrates, combined with relatively simple image analysis, provided growth data comparable to data obtained by turbidity measurements and by fluorescence microscopy. The method is carried out at the one-tenth millilitre scale. It thus represents a huge saving in laboratory facilities, consumables and labour, enabling a cheap and fast screening of growth substrates. In addition, the method avoids sampling and is non-intrusive. A digital camera with good resolution and an imaging software is all it requires, in contrast with the currently used methods involving expensive microscopy techniques. Furthermore, it may be used when cell clustering occurs, contrarily to optical density measurements.
Acknowledgement This study was supported by a post-doctoral grant (SFRH/BPD/14426/2003) awarded to Carla da C.C.R. de Carvalho by Fundac¸ a˜ o para a Cieˆ ncia e a Tecnologia, Portugal.
References Alcon, A., Santos, V.E., Casas, J.A., Garcı´a-Ochoa, F., 2004. Use of flow cytometry for growth structured kinetic model development. Application to Candida bombicola growth. Enzyme Microb. Technol. 34, 399 – 406. Bishop, P., Rittmann, B.E., 1995. Modelling heterogeneity in biofilms: report of the discussion session. Water Sci. Technol. 32, 263 – 265.
140
C.C.C.R. de Carvalho et al. / Journal of Microbiological Methods 60 (2005) 135–140
Davey, H.M., Kell, D.B., 1996. Flow cytometry and cell sorting of heterogeneous microbial populations: the importance of singlecell analyses. Microbiol. Rev. 60, 641 – 696. de Carvalho, C.C.C.R., da Fonseca, M.M.R., 2003. A simple method to observe organic solvent drops with a standard optical microscope. Microsc. Res. Tech. 60, 465 – 466. de Carvalho, C.C.C.R., da Fonseca, M.M.R., 2004. Principal component analysis applied to bacterial cell behaviour in the presence of organic solvents. Biocatal. Biotransform. 23, 1 – 12. de Carvalho, C.C.C.R., van Keulen, F., da Fonseca, M.M.R., 2000. Biotransformation of limonene-1,2-epoxide to limonene-1,2diol by Rhodococcus erythropolis cells: an introductory approach to selective hydrolysis and product separation. Food Technol. Biotechnol. 38, 181 – 185. de Carvalho, C.C.C.R., Parreþo-Marchante, B., Neumann, G., da Fonseca, M.M.R., Heipieper, H.J. (in press) Adaptation of Rhodococcus erythropolis DCL14 to growth on n-alkanes, alcohols and terpenes. Appl. Microbiol. Biotechnol. DOI:10.1007/s00253-004-1750-z. Donelly, C.W., Gaigent, G.J., 1986. Method for flow cytometric detection of Listeria monocytogenes in milk. Appl. Environ. Microbiol. 52, 689 – 695. Fry, J.C., 1990. Direct methods and biomass estimation. Methods Microbiol. 22, 41 – 85. Heldal, M., Norland, S., Riemann, B., Bratbak, G., 1994. Determination of bacterial cell number and cell volume by
means of flowcytometry. TEM and epifluorescence microscopy. J. Microbiol. Methods 20, 255 – 263. Kepner, R.L., Pratt, J.R., 1994. Use of fluorochromes for direct enumeration of total bacteria in environmental samples: past and present. Microbiol. Rev. 58, 603 – 615. Laplace-Builhe´, C., Hahne, K., Hunger, W., Tirilly, Y., Drocourt, J.L., 1993. Application of flow cytometry to rapid microbial analysis in food and drinks industries. Biol. Cell 78, 123 – 128. Paul, J.H., Myers, B., 1982. Fluorometric determinatiion of DNA in aquatic microorganisms by use of Hoechst 33258. Appl. Environ. Microbiol. 43, 1393 – 1399. Poglazova, M.N., Pomoshchnikova, N.A., Meisel, M.N., 1984. Determination of primary amines and protein using fluorescamine in suspensions of microorganisms. Microbiology 53, 20 – 23. Poglazova, M.N., Mitskevich, I.N., Kuzhinovsky, V.A., 1996. A spectrofluorimetric method for the determination bacterial counts in environmental samples. J. Microbiol. Methods 24, 211 – 218. Tyndall, R.L., Hand, R.E., Mann, R.C., Jernigan, R., 1985. Application of flow detection and characterization of Legionella. Environ. Microbiol. 49, 852 – 857. Wiegant, W.M., de Bont, J.A.M., 1980. A new route for ethylene glycol metabolism in Mycobacterium E44. J. Gen. Microbiol. 120, 325 – 331.