|Home | About | Journals | Submit | Contact Us | Français|
A novel, quantitative method for detecting poly-3-hydroxybutyrate (PHB) amounts in viable cells was developed to allow for high-throughput screening of mutant libraries. The staining technique was demonstrated and optimized for the cyanobacterium Synechocystis sp. strain PCC6803 and the eubacterium Escherichia coli to maximize the fluorescence difference between PHB-accumulating and control cells by flow cytometry. In Synechocystis, the level of nonspecific dye binding was reduced by using nonionic stain buffer that allowed quantitation of fluorescence levels. In E. coli, the use of a mild sucrose shock facilitated uptake of Nile red without significant loss of viability. The optimized staining protocols yielded a linear response for the mean fluorescence against (chemically measured) PHB. The staining protocols are novel methods useful in the high-throughput evaluation of combinatorial libraries of Synechocystis and E. coli using fluorescence-activated cell sorting to identify mutants with increased PHB-accumulating properties.
Metabolic engineering, the directed improvement of strains using modern genetic tools, has relied on rational and combinatorial approaches to meet its objectives. Rational approaches in metabolic engineering require a priori knowledge of the kinetics and regulatory structures of the product-synthesizing pathway. This approach is very effective but can be limited by the amount of required knowledge of the system. Combinatorial approaches do not require much prior knowledge but instead rely on methods for the rapid assessment of the phenotypes of a diverse mutant library to identify clones with improved properties. Inverse metabolic engineering (4, 10) embodies the essence of the combinatorial approach, since it consists of generating genomic perturbations, such as gene knockouts and overexpression mutants, and then screening them for the desired phenotype. High-scoring clones are then isolated, and the genomic perturbation is identified. This approach may also uncover fundamental information about the system, serving as a driver for follow-up rational modifications. Many genomic perturbations have been developed based on transposons, plasmid overexpression, gene shuffling, and other random mutagenesis methods (1, 2). An integral component of combinatorial approaches is a high-throughput screen for efficiently probing the library diversity.
Polyhydroxyalkanoates (PHAs) are a broad class of naturally occurring, commercially interesting thermoplastics (3). These plastics are attractive because they can be produced from renewable resources and are biodegradable. Poly-3-hydroxybutyrate (PHB) is the most abundant naturally occurring PHA and can be produced in a variety of microorganisms including chemotrophic and phototrophic bacteria. Despite the microbial diversity, detailed research has been focused on a few representative microbes such as Ralstonia eutropha or Synechocystis sp. strain PCC6803 (7, 23, 28). As well, after the successful cloning of the PHB biosynthesis genes, recombinant Escherichia coli has been an interesting model for PHB research (17, 18, 21).
Many rational approaches have been implemented to increase PHB productivity, including both environmental medium manipulations and genetic modifications (12, 14, 23). Basic data have been captured in models that computationally predict carbon flux distribution, specifically toward PHB synthesis in different media and genotypes (8, 27). Of several candidate species, recombinant E. coli has been well studied and is considered a primary candidate for industrial PHB production based on its high accumulation of PHB. Synechocystis is also interesting as a PHB producer because carbon dioxide is the sole carbon source, and genetic modification identified in Synechocystis may be applicable to PHB engineering in higher plants.
Prior PHB screening methods used Nile red, a dye that stains PHB and other neutral lipids in bacteria (16). These staining methods can be classified into two categories: (i) nonlethal and qualitative and (ii) lethal and quantitative. Nile red staining protocols that keep cells viable have been able to differentiate only non-PHB-producing from PHB-producing cells (22). Presumably, this is due to inefficient membrane permeation of Nile red in living cells. While this is helpful for identifying organisms that make PHB, it lacks the sensitivity necessary to detect the incremental improvements that one might expect in an engineered strain over a parental strain. Prior quantitative assessments of PHB based on Nile red fluorescence have involved various fixing steps (5, 6, 9, 15, 26). The fixing step, typically executed with an alcohol or acetone treatment, facilitates the permeation of the dye through the membrane. While this allows for accurate detection of PHB levels, the lethal nature of the protocol prevents its use in a screen for detecting mutants with improved PHB accumulation.
A staining protocol that can accurately measure PHB content in a combinatorial screen must allow the stain to enter the cell, specifically stain granules, and maintain high viability. Dye binding specificity (and the resulting fluorescence intensity) is a thermodynamic property and a function of the staining conditions (temperature, ionic strength, etc.). Cell permeabilization methods have been developed for other applications, such as extracting proteins from the outer membrane (25) or transforming bacteria with plasmids (20). Adapting such methods to transport Nile red across the membrane could allow efficient staining of the PHB granules. In this paper, we describe the development of two novel methods for detecting PHB based on Nile red fluorescence that both are quantitative and maintain cell viability. Each protocol can distinguish incremental differences in PHB appropriate for library screening in E. coli or Synechocystis.
Synechocystis sp. strain PCC6803 was grown at 30°C in BG11 medium (19). Cultures were grown in a light-tight incubator (E-36; Percival Scientific, Boone, Iowa) at 100 μmol of photons m−2 s−1 provided by cool white fluorescent lightbulbs. Cultures containing different amounts of PHB were prepared by growing them in BG11(P), a modified BG11 medium with 0.018 mM K2HPO4 (10% of the concentration in BG11), and/or supplementing the medium with 10 mM acetate for 8 to 12 days. Cells were cultured in Erlenmeyer flasks on a rotary shaker. A control strain in which PHB made up 0% (dry weight) of cells (0% PHB DCW) was made using a PHB synthase knockout, the phaESyn mutant strain (24).
E. coli (XL1-Blue; Stratagene, La Jolla, Calif.) transformed with a modified pJOE7 (13) plasmid was cultured at 37°C in Luria-Bertani (LB) medium containing 20 g/liter glucose and 25 μg/ml kanamycin. The modified pJOE7 was kindly provided by Anthony Sinskey; contains the genes phaAB from R. eutropha, encoding the β-ketothiolase and the acetoacetyl coenzyme A reductase, and phAEC from Allochromatium vinosum, encoding the two-subunit PHB polymerase; and encodes kanamycin resistance. As a no-PHB control, the same plasmid without the pha genes was also cultured. Optical density was used to track cell growth using an Ultraspec 2100pro spectrophotometer (Amersham Biosciences, Uppsala, Sweden). Synechocystis and E. coli were tracked by absorbance at 730 nm and 600 nm, respectively.
A Nile red (Sigma-Aldrich, St. Louis, MO) stock solution was made by dissolving the dye to 1 mg/ml in dimethyl sulfoxide (DMSO) unless otherwise noted. Three microliters of stock solution was added to 1 ml of staining buffer as indicated in the staining optimization. Fluorescence-activated cell sorting (FACS) was carried out on a FACScan (Becton Dickinson, Mountain View, CA) flow cytometer using the following settings: Synechocystis, forward scatter = E00, side scatter = 411, FL-1 = 582, and FL-2 = 551, and E. coli, forward scatter = E00, side scatter = 411, FL-1 = 582, and FL-2 = 535. Cells were excited with anair-cooled argon ion laser (488 nm), and FL-2 (585 nm) was used to detect Nilered fluorescence. Flow cytometry analysis was done on 50,000 cells using WinMDI 2.8 software.
Staining effectiveness was characterized by resolution,
where Mn is the geometric mean of the fluorescence distribution of n (n = 1 is the PHB-producing cell, n = 2 is the no-PHB control), δn is the standard deviation of the fluorescence distribution, and RS is a quantitative measure of the ability to differentiate two populations. Cell viability was assessed by the ratio of the CFU in the final stained preparation to cells from the medium.
PHB was analyzed as shown previously (24). More than 10 mg of cells was collected from culture by centrifugation (10 min, 3,200 × g). The resulting cell pellet was washed once with cold deionized water and dried overnight at 80°C. The dry cells were boiled in 1 ml of concentrated sulfuric acid for 60 min and then diluted with 4 ml of 14 mM H2SO4. Samples were centrifuged (15 min, 18,000 × g) to remove cell debris, and liquid was analyzed by high-pressure liquid chromatography using an Aminex HPX-87H ion-exclusion column (300 × 7.8 mm; Bio-Rad, Hercules, Calif.) (11). Commercially available PHB (Sigma-Aldrich, St. Louis, Mo.), processed in parallel with the samples, was used as a standard.
Synechocystis wild-type and phaESyn mutant cultures were grown for 7 days.
Cells were centrifuged (5 min, 3,000 × g), and resuspended to an A730 of 0.4 in 0.9% (wt/vol) sodium chloride solution. Three-microliter volumes of different concentrations of Nile red solution were added to 1-ml volumes of resuspended cells to final concentrations between 30 and 30,000 ng/ml. The mixture was incubated in the dark for 30 min and analyzed on the FACScan flow cytometer.
Deionized water and 0.9% (wt/vol) sodium chloride were used to resuspend the cells for staining to an A730 of 0.4. Three microliters of 10 mg/ml Nile red in DMSO was added to 1 ml of resuspended cells. The mixture was incubated in the dark for 30 min and analyzed on the FACScan flow cytometer.
Staining was performed as described above. After staining, cells were centrifuged (5 min, 3,000 × g) and resuspended to the same volume in 1% (wt/vol) sucrose, 1% (wt/vol) DMSO, phosphate-buffered saline, phosphate-buffered saline plus 1% (wt/vol) DMSO, 0.9% (wt/vol) sodium chloride plus 1% (wt/vol) DMSO, or deionized water. Cells were incubated in the dark for 30 min and analyzed on the flow cytometer.
E. coli XL1-Blue harboring the modified pJOE and the no-PHB control was cultured as described above.
Cultures were grown to stationary phase. Sucrose shock, isopropanol treatment, dimethyl sulfoxide treatment, and heat shock permeabilization methods were tested for resolution and viability after the shock as follows. Sucrose shock was carried out as shown previously (25). One milliliter of culture was cooled to 4°C for 10 min. The cells were then centrifuged (3 min, 3,000 × g, 4°C) and resuspended in 1 ml ice-cold TSE buffer (10 mM Tris-Cl [pH 7.5], 20% [wt/vol] sucrose, 2.5 mM Na-EDTA). The TSE mixture was incubated on ice for 10 min and then resuspended (3 min, 3,000 × g, 4°C) in 1 ml deionized water with 3 μl 10 mg/ml Nile red in DMSO. The solution was incubated in the dark for 30 min and analyzed on the FACScan flow cytometer. Isopropanol-shocked cells were prepared by centrifugation (3 min, 3,000 × g) and resuspension in 70% (wt/vol) isopropanol for 15 min. Cells were then centrifuged (3 min, 3,000 × g), resuspended in deionized water with 3 μl of 10 mg/ml Nile red in DMSO, incubated for 30 min in the dark, and analyzed on the FACScan flow cytometer. DMSO shock was performed by centrifuging (3 min, 3,000 × g) 1 ml of cell culture. Fifty microliters of 10 mg/ml Nile red in DMSO was added directly to the pellet. The pellet was quickly vortexed and diluted to 1 ml in water after incubation for 30 s. Cells were incubated for 30 min in the dark and analyzed on the FACScan flow cytometer. Heat shock was performed as described previously for preparation of competent cells (20). One milliliter of cells was cooled on ice for 10 min. Cells were then centrifuged (3 min, 3,000 × g, 4°C) and resuspended in 1 ml cold 80 mM MgCl2-20 mM CaCl2. The sample was then centrifuged (3 min, 3,000 × g, 4°C) and resuspended in 1 ml 0.1 M CaCl2 with 3 μl of 10 mg/ml Nile red in DMSO. Cells were heat shocked at 42°C for 90 s, incubated for 30 min in the dark, and then analyzed on the FACScan flow cytometer.
Cells were prepared by sucrose shock using TSE buffer with various sucrose concentrations (0, 5, 10, 15, and 20% [wt/vol]). Nile red at 10 mg/ml in DMSO was used for staining.
Cells were prepared by sucrose shock using 3 μl of different Nile red solutions to a final concentration between 30 and 30,000 ng/ml.
To maximize the resolution between PHB-producing strains and controls in Synechocystis, Nile red concentration and staining buffers were optimized to increase specific staining while minimizing nonspecific staining. A destaining step was introduced and evaluated, presuming that it would remove nonspecific stain from the cells. Figure Figure1A1A shows the effect of Nile red concentration on resolution for Synechocystis. Nile red at 3.3 μg/ml was found to yield the best resolving power. Nile red's fluorescence intensity is a function of the ionic character of the solution. Water and 0.9% (wt/vol) sodium chloride were examined as staining buffers to test the effect of solution ionic strength on Nile red staining and fluorescence. Water showed a much higher difference in cell staining than did the sodium chloride solution (Fig. (Fig.1B).1B). Several destaining solutions were examined to attempt to reduce nonspecific staining. While the sodium chloride-based buffers performed worse than the nonionic buffers, none of the destaining buffers significantly improved the resolution of the assay (data not shown).
The optimized protocol is as follows. Harvest cells by centrifugation (5 min, 3,000 × g). Resuspend cells to an A730 of 0.4 in deionized water. Add 3 μl of a stock solution of 1 mg/ml Nile red in DMSO to 1 ml of the cell suspension and incubate the mixture in the dark for 30 min. Analyze by flow cytometry immediately.
Following protocol optimization, cultures grown under different conditions were assayed by the Nile red staining protocol and chemical PHB analysis. As shown in Fig. Fig.2A,2A, the geometric mean of the flow cytometer measurement correlated very well with the analytical PHB measurements over a wide dynamic range of PHB concentrations in different growth media. Comparison of the Nile red fluorescence histogram for cells stained with 30 μg/ml Nile red in 0.9% (wt/vol) sodium chloride (Fig. (Fig.2B)2B) (protocol from reference 6) and the histogram for the optimized protocol (Fig. (Fig.2C)2C) demonstrated a significant improvement in the overall staining. Nonspecific staining in the overall population has been reduced by a shifting of the primary 0% PHB DCW peak to lower fluorescence, as well as a reduction of a secondary peak that was a result of strong nonspecific staining.
The optimized Synechocystis staining protocol was used as a starting point in investigating E. coli staining. Very poor staining of PHB was observed with this protocol, even in stationary phase, where PHB accumulates to a level of approximately 40% DCW. Assuming that the observed staining differences are due to greater dye permeation through Synechocystis cell membranes, methods to improve the dye permeation of E. coli were examined. Figure Figure3A3A shows the resolution achieved by different permeabilization methods on stationary-phase E. coli cells. Stationary-phase E. coli was used to ensure large differences in PHB. Of these, sucrose shock permeabilization gave the best resolution between high- and no-PHB strains, albeit with a very low viability of 0.022%.
To improve cell viability, the effect of varying the sucrose concentration used in the sucrose shock was studied. Exponential-phase E. coli, which did not have as large a difference in PHB accumulation, was found to be more sensitive to the sucrose shock and was examined as a worst-case scenario for viability. Figure Figure3B3B shows the tradeoff between higher viability and better staining. Overall, a 10% (wt/vol) sucrose concentration was found to yield staining characteristics similar to those obtained with a higher sucrose concentration, while maintaining a viability of 21%. Nile red concentration was varied to optimize resolution. A Nile red concentration of 33 μg/ml gave the best separation of the peaks (Fig. (Fig.3C).3C). This increase in optimal Nile red concentration over that for Synechocystis is most likely due to increased amounts of PHB in E. coli. Concentrations above this level were not soluble in the stock solution and gave considerable noise in the measurement (data not shown). Cell viability was not affected by Nile red concentration (data not shown).
To further improve the viability of the cells after the staining procedure, different resuspension buffers used after the TSE incubation step were evaluated. Water, LB, and 1 mM magnesium chloride had viabilities of 21%, 18%, and 48%, respectively. While even with the magnesium chloride only half the population survived, this only doubled the number of cells that would needed to be sorted in order to screen a library.
The optimized E. coli staining protocol is as follows. Cool cell culture to 4°C. Harvest cells by centrifugation (5 min, 1,000× g, 4°C). Resuspend cells to an A600 of 0.4 in 10% (wt/vol) sucrose TSE buffer and incubate them on ice for 10 min. Centrifuge (5 min, 3,000 × g, 4°C) and resuspend cells to the same volume in 4°C 1 mM MgCl2. Add 3 μl of a 1 mg/ml Nile red solution in DMSO to 1 ml of the cell suspension. Incubate the mixture in the dark for 30 min. Analyze by flow cytometry immediately.
Figure Figure4A4A shows the correlation between the PHB measurement obtained by the optimized E. coli Nile red staining protocol and that obtained by chemical PHB analysis. The data shown in Fig. Fig.4A4A correspond to cells harvested at different time points along a growth curve over a broad range of PHB amounts from exponential to stationary phase. Figures Figures4B4B and and4C4C show the dramatic improvement in staining achieved in the final protocol. As indicated in Fig. Fig.4B,4B, a large portion of the population was not stained in the original protocol but was stained after introduction of the sucrose shock step.
High-resolution FACS experiments can sort cells from a population only to the extent that the fluorescence level of the cell correlates with the phenotype that is being examined. In trying to develop a staining protocol that yields good correlation between fluorescence and, in this case, PHB levels, important factors are (i) promoting specific fluorescent molecule binding to the target (PHB) and (ii) allowing adequate access of the fluorescent molecule to the target.
In the Nile red optimization for PHB fluorescence, molecule specificity was addressed by varying the environment for staining. This was especially important in Synechocystis. Synechocystis has multiple layers of thylakoid membrane, which is used in its photosynthetic apparatus. These membranes provide a large area of lipid-like interfaces for nonspecific binding of Nile red. To minimize binding of the Nile red to the thylakoid membrane, the ionic strength of the medium and the dye concentration were examined. When the ionic strength of the staining environment was changed from a sodium chloride solution to deionized water, the Nile red stained the PHB more specifically. Additionally, the dye concentration also strongly affected the resolution of the assay. In E. coli, this was not as important. The effect of dye concentration increased, unlike in Synechocystis, where a maximum was observed. The ionic strength did not affect the resolution in E. coli (data not shown). These observations can be attributed to the lack of large membrane structures in E. coli.
Providing consistent access of the fluorescent molecule to the target is also necessary for determining a quantitative fluorescence level. In Synechocystis, the stain readily permeated the cell and stained the PHB granules. Figure Figure2B2B shows that even in the prior staining protocol, all Synechocystis cells were being stained. This should be contrasted with Fig. Fig.4B,4B, which shows that a large portion of the E. coli cells did not stain at all for PHB in the nonoptimized protocol. Synechocystis is a naturally competent cell and as such is able to take up DNA molecules readily. This may imply that the morphology of the Synechocystis membrane may allow it to take up Nile red more readily than E. coli, which is not naturally competent. To improve the dye transport across the E. coli cell membrane, competent-cell protocols and other permeabilization methods were attempted. Of these, sucrose shock permeabilized the cells in such a way that the Nile red could enter the cytoplasm and stain the granule.
While the E. coli cells could now take up the Nile red, most of the cells were killed in the process. Further optimization was required to increase the cell viability while retaining good staining properties. Adjusting the sucrose concentration and the buffers used improved the viability to 48%. This will allow an adequate efficiency for screening mutant libraries by FACS.
To validate the use of resolution (equation 1) as a metric for optimizing the protocol and to estimate the accuracy of the Nile red fluorescence, the geometric mean of the fluorescence distribution was compared to a chemical PHB measurement of the culture. As there is presently no validated method for measuring PHB levels at the individual cell level, population average measurements, such as the geometric mean of fluorescence and the whole-culture chemical PHB measurement, were required to assess the quantitative accuracy of the staining protocols. The correlation between fluorescence and PHB content was greatly improved over that of initial staining experiments (data not shown) due to the improved staining of PHB granules and reduction in nonspecific staining. The estimated error of prediction of PHB content from the geometric mean of fluorescence was ±1.2% PHB DCW and ±4.5% PHB DCW for Synechocystis and E. coli, respectively (95% confidence interval). From this, it can be inferred that the PHB levels on the single-cell level can be estimated accurately based on the fluorescence measurement.
These protocols will allow single-cell measurements of PHB levels in Synechocystis and E. coli to such a level of precision that mutants with incrementally increased PHB accumulation can be sorted from the library and characterized. Using FACS, 10 million cells can easily be assayed in less than 1 h. While there will be a loss due to nonviable cells in the E. coli system, this loss does not prohibit the assay from screening genome-scale libraries. As well, multiple cells of the same genotype will be present due to growth, increasing the likelihood of each library variant being screened.
Biological noise will most likely contribute false positives to the screen. Inherent in all single-cell measurements is the cell-to-cell variation even in a clonal population. This is evident in Fig. Fig.2C2C and and4C.4C. For the positive controls, a clonal population has a 10-fold difference in fluorescence within the population. This variation in PHB content will result in false positives being sorted as high-PHB clones, while their average PHB content may be less.
The application of sucrose shock to allow E. coli to take up Nile red is generalizable to other bacteria and other small-molecule dyes which do not permeate the membrane. By using such permeabilization methods to allow impermeant fluorescent dyes to enter the cytoplasm, the number of phenotypes that can be screened in a high-throughput fashion can be significantly increased. This will enable new fluorescence-based combinatorial screens for other phenotypes where high-throughput screens do not currently exist.
This work was supported by Department of Energy grant DE-FG02-94ER14487 and National Science Foundation grant 0331364.
Acknowledgments to Mark Olsen for his expertise in staining protocols and flow cytometry.