Variances in cellular sedimentation behavior as an effective enrichment method of hydrocarbon-overproducing Micrococcus luteus strains

Biosynthesized hydrocarbons are an attractive target for microbial metabolic engineering, because these molecules can be very similar or identical to the main components of the petroleum-derived fuels and are, thus, fully compatible with the existing infrastructure [1–4]. Several bacterial pathways have been discovered in the last years which lead to different types of hydrocarbons, among them is the cyanobacterial medium-chain alkane pathway [5], or the olefin biosynthesis pathway, which was first elucidated in Micrococcus luteus [6] and was later found to be present in various bacterial lineages [7]. The development of future bacterial hydrocarbon production will depend on the availability of tools to perform rapid genetic engineering as well as on methods to evaluate the hydrocarbon content in the engineered strains. Classically, microbial hydrocarbons and other fatty acid-derived fuels are measured by solvent extractions of bacterial cultures followed by gas chromatography–mass spectrometry (GC–MS). Although this is the most accurate and reliable method to measure such compounds, it is not well suited for high-throughput settings, for example, in a screening of a random mutagenesis library. For high-throughput assays, the fluorescence emitted by the hydrophobic dye Nile Red has been established as a convenient proxy for the amounts of hydrocarbons and ketones in bacteria [8] and Nile Red-based screenings have been employed in the isolation of free fatty acid overproducing mutants of Escherichia coli [9]. Classical screening approaches, like the Nile Red screening above, rely on the isolation and probing of single colonies (or liquid cultures). The present study describes the relation we observed between cell buoyancy and olefin production in Micrococcus luteus and the use of this phenotype to simply and efficiently separate cells from a mixture based on their hydrocarbon content. Selective enrichment of cells from mixed populations based on their size has been demonstrated before in continuous fermentations in a specially designed fermenter [10, 11]. With the population heterogeneity in hydrocarbon-producing M. luteus cells described by us, it may be feasible to apply similar fermentation strategies to increase productivity.

During our work on engineering the olefin biosynthetic pathway in Micrococcus luteus, we repeatedly observed that the cells of liquid cultures expressing different amounts of olefins show distinct sedimentation behavior. For example, when cell suspensions of stationary phase-grown M. luteus trpE16 (parental strain, [12]), M. luteus ΔoleABCD:kan (a non-hydrocarbon-producing mutant of trpE16) and M. luteus ope (an olefin-overproducing variant of trpE16) were left undisturbed for 4 h, a clear difference in the speed with which the cells sedimented could be observed (Fig. 1a). These differences were not due to different growth stages of the three strains, as they showed almost identical growth kinetics under these conditions (Fig. 1b). Also, there were no noteworthy differences in the microscopic appearance of the cells, i.e., regarding cell size and cell aggregation behavior (see below). We then asked if this settling phenotype can be sufficient to separate hydrocarbon-producing from non-producing cells in a mixed population simply by collecting the upper phase of a cell suspension after letting it to stand undisturbed for several hours. For this, cell suspensions of a mix of two different M. luteus strains with distinct olefin content were prepared, where the mixes contained approximately equal amounts of the two strains. In these cell mixes, one of the strains carried a chromosomally integrated kanamycin resistance marker, which permitted us to easily determine the ratio of the strains in the samples by plating on agar plates with and without antibiotic. From a glass tube with a 10-ml cell suspension mix left standing unmoved at 20 °C, 0.1 ml of samples from near (about 5 mm) the top was collected at different time points and dilutions of the samples were plated on LB plates with and without kanamycin to determine the relative abundance of the kanamycin-resistant cells. When olefin-non-producing cells were mixed with either the wild type or the M. luteus ope strain, the fraction of olefin-non-producing cells in the upper phase decreased dramatically, leading to a more than 1000-fold enrichment of olefin-producing cells in the upper phase after 20 h (mixes A and B in Fig. 2a). Control experiments, performed with mixtures of the wild type and an olefin-overproducing strain carrying a kanamycin marker (trpE16 and ope:kan, mix C in Fig. 2a), showed that this separation was not affected by the antibiotic resistance gene. Another control involved mixing of the wild type and a strain with a kanamycin insertion not affecting the olefin production (trpE16 and Δ02970:kan, the latter being a deletion strain for a putative transcription regulator, mix D in Fig. 2a), where the strains ratio was deliberately set at ~ 5 × 10⁻² at the beginning to be able to follow changes in both directions. This control experiment also supported the hypothesis that the enrichment we observed in the other mixes was olefins dependent. Moreover, we observed that when the wild type was mixed with the olefin-overproducing strain, an efficient enrichment of the latter strain in the upper phase of the culture was possible (mix C in Fig. 2a, ope:kan abundance changes from 20% to 100% after 20 h).

We further confirmed that such a simple sedimentation approach can indeed separate hydrocarbon-producing from non-producing cells of M. luteus by measuring the olefin content in hexane extracts, obtained from fractions of a mixture of the two strains ope and ΔoleABCD:kan (Fig. 2b). The two strains were grown separately to stationary phase in rich medium and equal amounts of cells were mixed (in a total volume of 800 ml) in a cylindrical glass vial and were let to stand undisturbed for 4 h. Samples were collected from the pure cultures, from the freshly mixed cell suspension (sample ‘1:1 Mix A’), and three samples from the mixed cell suspension after standing for 4 h, corresponding to the upper phase (Fr. 1), lower phase (Fr. 2) and the cell pellet, (Fr. P) as shown in Fig. 2b. The cells from all samples were collected by centrifugation and after normalizing by optical density were subjected to extraction with hexane. Quantitative GC–MS analysis of these samples clearly showed that the upper phase contained most of the olefins, which means most of the olefin-producing cells; while in the lower phase and the cell pellet, which formed after several hours of incubation, almost no olefins could be detected.

Among several possible explanations of the observed phenotypic differences between olefin-producing and non-producing cells, the most obvious one is a difference in buoyant cell density. To test this, we performed buoyant density determination of the three M. luteus strains (trpE16, ΔoleABCD:kan and ope) by equilibrium centrifugation in Percoll gradients. Several independent cultures of each of the three strains were analyzed by centrifugation of the cells in in situ-formed Percoll gradients (initial density adjusted to 1.09 g × ml⁻¹ in 0.15 M NaCl, centrifugation at 23,000×g for 30 min at 20 °C). A representative result from these measurements is shown in Fig. 3. Consistent with the observed differences in sedimentation, the buoyant densities of the examined strains showed small but well-detectable differences, the olefin-overproducing strain showing the lowest measured density. Indeed, it can be expected that the micrococcal olefins, which make up about 0.2% of the total dry cell weight in the wild type M. luteus strain ([13], and our own measurements), would have a measurable impact on the cell density. Considering an average density of 0.79 g × ml⁻¹ for the olefins of Micrococcus (consisting mainly of C27 to C30 mono-alkenes), and a density of 1.0923 mg × ml⁻¹ for the trpE16 strain, a fivefold increase in the cellular olefins (from 0.2 to 1% of the CDW as is the case in the ope strain, measured by GC–MS) would theoretically lead to a density of 1.0898 g × ml⁻¹, which is in a remarkable agreement with the actually measured average density of the ope strain of 1.0892 g × ml⁻¹ (Fig. 3).

Another interesting observation from the Percoll gradient centrifugation experiments was the broader density distribution of the cells of the olefin-producing strains in comparison to the cells of the olefin-deficient knockout strain, which can be seen as broader, less compact banding of the trpE16 and ope cells in Fig. 3a. A representation of the “compactness” of the cell banding for the three Micrococcus strains, obtained by plotting the optical density (which corresponds to cell density) of scanned images of the centrifuge tubes along the length of the tube, is shown in Additional file 1: Figure S1. This observation suggests a more heterogenic population (in terms of density) of cells in the olefin-producing strains compared to the deficient one. We are currently investigating if this density heterogeneity is caused by heterogeneity in the olefin content of the producer cells.

In the equilibrium density gradient centrifugation technique (isopycnic centrifugation), the cells are separated solely on the basis of differences in density, irrespective of size, which can play a role only before the equilibrium is reached [14]. The buoyant density differences between olefin-producing and non-producing cells which we observed must, therefore, be associated exclusively with the cell density, as the Percoll centrifugation experiments were performed to equilibrium. We corroborated this with microscopy observations and measurements of the average cell sizes of the three strains used in the Percoll experiments, e.g., E16, ope and ΔoleABCD:kan. No significant differences in the cell perimeter were observed for the three strains when they were grown to the stationary phase in rich medium (Additional file 2: Figure S2).

It is not clear if the small differences in density we measured by centrifugation in Percoll gradients are the sole reason for the pronounced phenotypic differences (rate of sedimentation) between olefin-producing and non-producing Micrococcus cells. Irrespective of the exact physical basis for the different behavior of olefin-producing and non-producing cells, the simple separation method described here can be used in various mutagenesis and enrichment protocols for isolating strains with increased olefin content. For example, it should be possible to enrich for olefin-overproducing mutants from a pool of randomly mutagenized M. luteus cells by repeated separation and outgrowth of the cells from the upper phase after settling. Also, it may be possible to accelerate the sedimentation rate by the “Boycott effect”, e.g., using inclined vessels [11].

It would be interesting to investigate if the olefin content-associated phenotypic differences which we describe for Micrococcus can be found and utilized also in other microorganisms with a natural or engineered hydrocarbon-producing capability.

In summary, we show that cultures of M. luteus cells show distinct sedimentation behavior depending on the amounts of alkenes produced. Equilibrium centrifugation experiments in Percoll gradients of M. luteus mutants suggest that buoyant cell density is the primary cause for the observed differences in sedimentation behavior. We demonstrate how this phenotypic trait can be used to simply and efficiently fractionate M. luteus cells based on their hydrocarbon content. This approach can be useful in the isolation of olefin-overproducing mutants of M. luteus and probably other hydrocarbon-producing species.