Cultivation conditions. Wild-type *W. eutropha* H16 was cultivated with aeration at 30°C. Gentamicin was included in all growth media, except when PHB utilization was being measured. A single colony from a dextrose-free tryptic soy broth (TSB) (Becton Dickinson Microbiology Systems, Cockeysville, MD) plate was cultivated in 5 ml of TSB to saturation (~40 h), at which time 2 ml was transferred into 100 ml of TSB in 500-ml baffled flasks and grown for 24 h. The doubling time of *W. eutropha* H16 in TSB is between 3 and 4 h. Cells harvested by centrifugation were washed and transferred into 200 ml of TSB or 200 ml of PHB_{P} (minimal medium supplemented with 1% fructose and 0.01% [wt/vol] ammonium chloride) in 1-liter baffled flasks to obtain cultures with an initial optical density at 600 nm of 0.5. For cells grown under TSB conditions, 5 ml of cells was removed at 4 and 24 h for TEM analysis. For cells grown in PHB_{P}, 5 ml of cells was removed from the culture at 2.5, 5, 9, 24, and 73 h. In all cases, cells were immediately fixed for TEM studies. For PHB utilization, 100 ml of cells grown in PHB_{P} for 73 h was harvested, washed with 0.85% (wt/vol) saline, and transferred into 200 ml of PHB utilization medium (PHB_{U}) (minimal medium supplemented with 0.5% [wt/vol] ammonium chloride). Samples were harvested at 48 h for TEM analysis.

Reagents. All TEM reagents were purchased from Electron Microscopy Sciences (Hatfield, PA).

Fixation. Five milliliters of the cell culture in TSB, PHB_{P}, or PHB_{U} at various times was transferred to a 15-ml Falcon tube containing 5 ml of fresh fixative solution (2% [vol/vol] glutaraldehyde, 3% [wt/vol] paraformaldehyde made fresh, 5% [wt/vol] sucrose, and 0.1 M sodium cacodylate buffer, pH 7.4). After 5 min of manual mixing, the cells were spun down at 5,000 rpm for 10 min using a bench centrifuge. The supernatant was removed, and an additional fresh 10 ml of the fixative solution was added to resuspend the cell pellet. After 1 h of incubation at room temperature with occasional manual mixing, the cells were pelleted again. Sodium cacodylate buffer (0.1 M at pH 7.4; 1.5 ml) was added to resuspend the pellet. The cell suspension was transferred to an Eppendorf tube (1.7 ml), and after 2 min of mixing, the cells were spun down at 9,000 rpm for 1 min in a minicentrifuge. The supernatant was removed, and the pellet was washed three times with 1.5 ml of 0.1 M sodium cacodylate buffer. The pellet was dislodged from the bottom of the Eppendorf tube to ensure good washing each time.

The cell pellet was then fixed with 1.5 ml of a 1% osmium solution prepared by mixing 1.25 ml of 4% osmium tetroxide (OsO_{4}) (Electron Microscopy Sciences), 1 ml of 0.1 N HCl, 1.75 ml of distilled H_{2}O, and 1 ml of acetate-Veronal stock (1.2% [wt/vol] anhydrous sodium acetate and 2.9% [wt/vol] sodium barbituate [Veronal] in distilled H_{2}O). The pellet was dislodged from the bottom of the tube and incubated in the osmium solution for 1 h.

The fixed cells were then pelleted, and 1.5 ml of a third fixative solution, the Kellenberger uranyl acetate solution (0.5% [wt/vol] uranyl acetate in Veronal-acetate buffer) was added. The pellet was initially washed with this solution briefly (<1min) and then incubated in 1.5 ml of fresh solution overnight in the dark.

Dehydration. Upon the completion of uranyl acetate staining, the pellet was rinsed with ~1.5 ml of distilled H_{2}O quickly after being dislodged and was then pelleted by centrifugation. To dehydrate the cells, they were subjected to increasing amounts of ethanol. The pellet was placed first in 50% (vol/vol) ethanol-water for 10 min and subsequently in 70% (vol/vol) ethanol for 10 min, 95% (vol/vol) ethanol for 10 min, and then three times in 100% ethanol for 15 min. The cells were spun down after each ethanol treatment to remove the supernatant, and the pellet was dislodged during each incubation to ensure homogeneous dehydration. The pellet was further dehydrated in 50% (vol/vol) ethanol-50% (vol/vol) propylene oxide for less than 5 min before it was transferred to 100% propylene oxide (~1.5 ml). After 5 min, the pellet was then placed in 50% propylene oxide-50% low-viscosity embedding resin (containing vinyl-4-cyclohexene dioxide, DER 736 resin, nonenyl succinic anhydride, and 2-dimethylaminoethanol; these components were mixed in proportions to obtain hard blocks when the instructions of a Spurr kit were followed) and rotated on a rotator for 12 h.

Embedding. All of the pellet was transferred into ~1.5 ml of 100% low-viscosity embedding resin and placed under a vacuum in a desiccator for 4 h. This process was repeated at least three times with fresh embedding resin. The pellet was then cut into small sections randomly and embedded in beam capsules containing 100% low-viscosity embedding resin. The beam capsules were then placed at 60°C overnight to allow embedding.

Sectioning and microscopy. Ultrathin sections (thickness, ~70 nm) were cut with a Reichert Ultracut E microtome using a Diatome diamond knife. The sections were picked up with 200-mesh nickel grids coated with Formvar (0.3% [wt/vol] dissolved in ethylene dichloride) and a layer of carbon. Serial sections were also prepared for wild-type *W. eutropha* cells grown in PHB_{P} for 5 h. The thickness of each section was ~70 nm. The sections were examined using a Philips EM410 or JOEL JEM-1200EXII electron microscope at 80 kV. For each condition, images were recorded on film at high and low magnifications.

TEM image data analysis: calculation of the average cell volume at 5 h of the wild-type strain grown in PHB_{P}. *W. eutropha* cells appear to be rod shaped under a light microscope, and all calculations of volumes are based on this premise. Thus, the volume of a cell has been approximated by using the equation describing the volume of a cylinder: *V*_{C} = π*d*^{2}*h*/4, where *V*_{C} is the volume of a cylinder or a single cell, *d* is the diameter of the cylinder or the width of the cell, and *h* is the height of the cylinder or the length of the cell. Serial sections (~70 nm; 5 h in PHB_{P}) were required to obtain the actual length and width of a cell by selecting the longest and widest cell profiles, respectively, in images. Here, cell profiles refer to the cross section of cells resulting from a single cut. Long cell profiles that did not change length and angle from one section to the next were measured with a ruler to determine the length (*h*); similarly, wide cell profiles that did not change width and angle from one section to the next were measured to determine the width (*d*). All measured values were corrected by the magnification factor. The average cell volume for this sample is referred to here as *V*_{C}_{5h}. This analysis assumed that the cells were uniform in size at this time, which was not the case. At 5 h under the growth conditions used, the cells were dividing. Since the analysis involved examination of successive sections for the longest cells which were on the verge of dividing, *V*_{C}_{5h} was the average cell volume at this stage. The average cell volume of freshly divided cells at 5 h was assumed to be one-half of *V*_{C}_{5h}. Thus, as noted in Table , a range of volumes is reported here for the times at which cells were dividing. In the stationary phase (TSB at 24 h and PHB_{P} at 9 to 73 h), the cells were assumed to be uniform.

| **TABLE 1.**Estimated average cell volume and total surface area of granules per *W. eutropha* H16 cell determined by the stereology method and reported as a function of time and cultivation conditions^{a} |

Calculation of the area of cell profiles on two-dimensional (2-D) images using unbiased stereology at different times. The Cavalieri point counting method was used to calculate the area of cell profiles (

5,

26). A multipurpose test system was used. The probe contained parallel lines (also called test lines) spaced 19.85 mm apart and points spaced 39.7 mm apart evenly on each line (http://web.mit.edu/biochemistry/PHB_Supp_1.pdf). The probe was overlaid randomly on a TEM image containing cell profiles. The number of points (

*P*) that hit cell profiles was then tabulated. This process was repeated with images of other random sections of the same sample until enough points were obtained so that the coefficient of error was less than 10%. The coefficient of error, also known as sampling error, is defined as follows: (standard deviation/mean)/(sample number)

^{1/2}, where the sample number is equal to the number of images used for the study. Typically, a minimum of 100 to 200 points is needed. The area of the cell profiles covered in all images was calculated using equation

1 (

16,

33).

where

*A*_{CP} is the total area of the cell profiles on all images, ∑

*P*_{CP} is the number of points hitting the cell profiles, summed over all images, and

*a*(

*p*) is the area per point, the product of the distances between points in the

*x* and

*y* directions (19.85 mm by 39.7 mm). Since all images were magnified,

*a*(

*p*) was corrected by the magnification factor.

Calculation of the average volume of cells at each time using *A*_{CP} of the corresponding sample and *V*_{C}_{5h}. The Delesse principle states that the two-dimensional (2-D) areas of profiles of tissue components are related to the three-dimensional volumes occupied in space by these components, assuming random distribution and random orientation of components (

6). The relationship between area and volume is shown in equation

2.

where

*A*_{A} is the area fraction and

*V*_{V} is the volume fraction. Both

*A*_{A} and

*V*_{V} are further defined as shown in equation

3.

where

*A*_{obj} is the area of the object of interest on flat images, which in our case was equal to the area of total cell profiles (

*A*_{CP}) described above;

*A*_{ref} is the area of the reference space that contained

*A*_{obj}, which can be obtained by counting the total number of points on the point-grid probe used to obtain

*A*_{CP} and multiplying the total number of points by the area per point (equation

1);

*V*_{obj} is the volume of the object of interest, which in our case was the total volume of the cells observed in images from which

*A*_{obj} was measured; and

*V*_{ref} is the reference volume that contained

*V*_{obj}. Since

*A*_{CP} is a function of ∑

*P*_{CP} (equation

1), a value that is highly dependent on the number of observed cell profiles present on TEM images which are all taken at random, a normalization procedure is necessary in order to compare

*A*_{CP} values calculated for samples collected at each time point. Therefore, the total number of cell profiles (

*N*_{CP}) in all the images measured is determined. In order to avoid overcounting, a counting rule is applied. The cell profiles that lie within the reference area and only those that are on the top and at the left edge of the reference area are counted. Note that

*N*_{CP} only refers to the total number of cell profiles that are observed and therefore contribute to the measurement of

*A*_{CP}. Therefore, the

*A*_{CP}/

*N*_{CP} values obtained from different samples are indicative of the trends in volume change. When parameters of separate samples, such as 2.5-h and 5-h PHB

_{P} samples, are compared, the relationship among the

*A*_{CP}, volume of cell profiles (

*V*_{CP}), and

*N*_{CP} when the thickness of each section is the same can be found in the equation

4.

Notice that we did not chose to cancel out the

*N*_{CP2.5 h} and

*N*_{CP5h} terms in this equation. This is because the terms in the parentheses (

*V*_{CP}/

*N*_{CP}) now represent the actual average volume of a cell (

*V*_{C}). This representation is valid only when the number of cell profiles is equivalent to the number of cells present in the reference space, which is true in our system since a cylindrical cell can appear only once when it is sliced from any angle. The only exception is when the cells are dividing. Since the middle of a cell pinches in, sectioning the cell longitudinally on the edge could result in two profiles. However, the probability of this is low (≤2 or 3% for ~100 cell profiles) (data not shown). In addition, cell division does not occur in PHB

_{P} at 9, 24, and 73 h. Since we obtained the actual average volume of a cell in PHB

_{P} at 5 h using images of serial sections as described above, knowing

*A*_{CP2.5 h},

*A*_{CP5h},

*N*_{CP2.5 h}, and

*N*_{CP5h} allowed us to calculate the approximate average volume of a cell in PHB

_{P} at 2.5 h. Similarly, an estimate of the average volume of a cell from other samples could be obtained using this method. Equation

4 is valid when the thickness of a section is the same for different samples, which was the case for this study.

Calculation of the total surface area of granules per cell using unbiased stereology. The multipurpose test system (http://web.mit.edu/biochemistry/PHB_Supp_1.pdf) was again overlaid on TEM images of the same sample randomly (all images were the same magnification). The size of the image defined the reference space. Since all images were approximately the same size, the total reference space was equal to the number of images times the size of an image. This time, instead of counting points, intersections (

*I*) were counted when the test lines crossed the surface of granules. Again, a large enough number of intersections had to be counted so that the sampling error was less than 10%. The surface density of the granules (

*S*_{V}), defined as the total surface area of granules per total reference space volume, could be calculated using equation

5 (

30).

where ∑

*I*_{G} is the number of intersections of granules with test lines, summed over all images, and ∑

*L* is the total length of test lines summed over all images. Again, the length had to be corrected for magnification. Note that the units for

*S*_{V} are length

^{−1} (e.g., μm

^{−1}). The derivation of this equation has been described by Elias et al. (

11; also see ). Recall that

*V*_{V} (

*V*_{obj}/

*V*_{ref}) is the total volume of observed cell profiles in the total reference space, which in our case was obtained from the same images. Therefore, dividing

*S*_{V} by

*V*_{V} gives the total surface area of granules in a unit volume of cells. The total surface area of granules in one cell (

*S*_{G}) can be obtained using the following equation.

Again, this analysis was applied to images of all our samples.

Measurement of size distribution of cell and granule profiles on 2-D images. For all wild-type samples, the major and minor axes of all cell and granule profiles were measured for images recorded at the same low magnification (primary magnification, ×3,000) so that a large field of cells could be sampled. Measurements were made by using the Scion Image for Windows software from Scion Corporation. These measurements were confirmed by random manual measurements with rulers.