## Cellular morphology changes during a growth curve

Figure 1. AFM topographs displaying bacterial morphological changes during a growth curve. All images are 15 µm x 15 µm and were obtained at population age A) 2 hours, B) 4 hours, C) 6 hours, D) 8 hours and E) 11 hours after inoculation. The only image post-processing performed was a simple first order plane fit. The boxed cells in each image are described in the text and presented in Fig. 4 in three-dimensional close-ups. Panel F shows cross-sections from panels A and C in the upper and lower graphs respectively. These cross-sections demonstrate the three-dimensional nature of AFM data and illustrate how cell length is measured. The star on an image and its corresponding cross-section are at about the same position. For the upper graph in Panel F, the cell length is defined as the distance between the two local minima at the cell’s end, as marked by the dashed lines, yielding a length of 1.14 µm for this cell. For the lower graph in Panel F, the local minima differ significantly in height, so here the cell length is defined as the distance between the shallowest local minimum and the equal height location on the opposite side of the cell, as marked by the dashed lines, yielding a length of 2.14 µm.

Both within and between the images, the panels in Fig. 1 show a striking amount of morphological heterogeneity. Cells from all time points (Panels A-E) contain both round and rod-shaped cells, while cells from later times also contain filaments (Panels C and E), with a particularly long filament boxed in Panel E. In addition, there is a striking amount of heterogeneity among the surface features of the cells. Some cells are smooth across their top surface (see boxed cell in bottom center of A), while some have small indentations towards their poles (see boxed cells at bottom left of C). Some cells appear sunken in throughout their centers (see boxed cells in D), while some cells are asymmetrically sunken at their midline (see boxed cells in B). Other species of bacteria also have some of these features when imaged with AFM (14, 16–21).

The growth curve corresponding to Fig. 1 is given in Fig. 2D, with the growth rate given in 2E. These figures show that the population is in exponential phase from 2–5 hours after inoculation. A transition from exponential to stationary phase occurs from 5–7 hours after inoculation, during which time the population is still increasing, but at a decreasing rate. Stationary phase begins 7 hours after inoculation at which time the growth rate is close to zero and remains that way going forward. Vertical gray lines on Fig. 2 delineate these three time periods.

Figure 1 shows a clear progression of shape from one panel to the next as cells go through the growth phases identified in Fig. 2D and E. Panels 1A and B, taken of exponential phase cells, show spherical cells, coccobacilli, and short rods. During the exponential to stationary phase transition at 6 hours after inoculation, the cells become longer rods as shown in panel C, and many septa are observed, an example of which is boxed at the left middle of the panel. In panels D and E, which derive from stationary phase proper, we see a decidedly narrower population of primarily rods.

In short, Fig. 1 illustrates an array of morphological heterogeneity that underlies the growth curves of Fig. 2D. To investigate further the interrelationship between morphology and growth phase, we quantified and categorized cellular size and morphology as described in the following paragraphs.

## Atomic force microscopy imaging and quantification

Atomic force microscopy (AFM) is a tool well suited to studying bacterial shape (10, 11, 13). Its resolution is limited by the sharpness of the tip of the cantilever used to take the image, which is typically of order 10 nm. Since usual bacterial dimensions are of order 1 µm, AFM can easily resolve small changes in bacterial shape. In addition, AFM produces true three-dimensional images as shown in Fig. 1F, which shows a cross section of cells from panels 1A and 1C. Finally, preparing cells for AFM imaging is straightforward.

To make a series of images as in Fig. 1, we sample a single culture at regular intervals to make samples and then image by AFM (see Methods for details). To quantify cell size, we image each sample in several disparate locations to ensure we see its full diversity. For each of the resulting images, we randomly choose 50 cells. For each cell, we measure its length and width and classify it according to its surface features. Figure 1F shows how individual cell measurements are made using a cross-section from 1A and 1C. In brief, we simultaneously use a cross-section and its corresponding image to locate the local minima that define the edges of the cells (as indicated by vertical dashed lines on the cross-sections). The distance between the minima is defined as the length or width, with length always the larger of the two distances.

Using thousands of individual cells’ measurements, in Fig. 2A, B and C we plot the progression of bacterial shape as a function of time since inoculation. All the data in these panels comes from a single time series, but the same morphological trends were evident in other independent time series. For the time series data presented in Fig. 2 we assessed three to four 30 µm X 30 ∝m images per hour, taken on the hour starting 2 hours and finishing 12 hours after the inoculation of the broth, for a total of 38 images and 1900 cells assessed. With this data in hand, we are now in a position to look quantitatively at the morphological trends underlying the growth curve.

## Interrelationship of growth curve and morphology changes

Figure 2A, B, and C show the changes of the average (red circles), median (black squares), and distribution (gray bars) of cell measurements during the growth curve shown in 2D with corresponding growth rate shown in 2E.

We focus first on cell length in Fig. 2A. The average length of the population is 1.91 µm for cells in exponential phase, defined to be from 2–4 hours after inoculation. In hour 5 just as the growth rate begins to decline, there is a rapid increase in average cell length. The average length for cells 5–7 hours after inoculation is 2.65 µm, so this represents a 39% increase in length over exponential phase cells. In addition, there is a diversification of lengths indicated by a doubling in size of the cell-length distribution gray bars. This diversification can be seen in more detail in the histogram of lengths shown in Fig. 3. Comparison of the distribution of length in exponential phase (Fig. 3G) to the distribution of length in the transition time (Fig. 3H) shows that for cells in exponential phase the distribution is bell shaped with weight approximately between 1.0 and 2.5 µm and small number of cells in the higher length tail, whereas for cells in transition, the histogram has shifted to being fairly flat between 1.0 and 3.4 µm.

Returning to Fig. 2A, during the transition, the average cell length remains high and highly diversified as the growth rate falls through hour 7. In stationary phase starting in hour 8 and persisting through hour 12, the average cell length falls to 1.75 µm, similar to exponential phase cells. Likewise, a comparison of the length histograms in Fig. 3G, H and I shows that the distribution in stationary phase resembles that of exponential phase, being bell shaped with weight approximately between 1.0 and 2.5 um.

We now consider the changes in cell width in Fig. 2B. In hours 2–4 in exponential phase cell width averages 1.13 µm. Cell width remains at about this value even as the growth rate begins to fall in hours 5 and 6. In hour 7, just as the growth rate reaches zero and stationary phase begins, the average width begins to decrease, settling in hour 8 and beyond at an average value of 0.87 µm which is 30% lower than its exponential phase value.

Using individual cell lengths and widths, it is possible to calculate the surface area to volume (SA/V) ratio for each cell and hence the average SA/V for all cells at a given time point. The changes in surface area to volume (SA/V) ratio are given in Fig. 2C. In hour 2–4 in exponential phase the average SA/V is 4.68 µm− 1. It drops in hour 5 due to the dramatic increase in cell lengths, and then between hours 5–8 during the transition from exponential to stationary phase, SA/V increases steadily. In stationary phase proper in hour 8 and beyond, the SA/V ratio settles in at an average value of 5.91 µm− 1 which is 26% higher than its exponential phase value.

In Fig. 3 we consider how the width and surface area to volume of cells vary with their length in the three time periods during the growth curve. To make this figure, we use cell length to bin all the cells in a given time period. For each length bin, we plot the average width (Fig. 3ABC) and the average surface area to volume (Fig. 3DEF). Figure 3GHI gives the counts for each length bin. Since cells lengthen as they age in preparation for cell division, it is useful to think of the horizontal axis of the plot as indicating increasing cell age.

From Fig. 3A, cells in exponential phase with lengths less than about 2.1 µm (vertical dashed line) are narrow with the width increasing as cells get longer. Cells with lengths longer than 2.10 µm have a more constant width with an average of 1.20 µm (horizontal dashed line). The trends for width vs. length are similar for cells in transition (3B) and stationary phase (3C); however, the average width approached decreases to 1.10 µm and 0.92 µm respectively for lengths longer than 2.10 µm and 1.60 µm in transition and stationary phase respectively. Note that for stationary phase, the asymptotic width is reached at a shorter length than in exponential phase or transition.

We now consider the SA/V as a function of length as shown in Fig. 3D, E and F. In exponential phase and during transition the SA/V is decreasing for lengths less than about 2.1 µm (vertical dashed line), whereas above 2.10 µm it reaches an asymptotic average value of 4.09 µm− 1 and 4.30 µm− 1 respectively. In stationary phase, the SA/V is decreasing at all lengths with a definitive inflection point at a length of about 1.6 µm. As was also see in Fig. 2C, the SA/V is higher in stationary phase than during the other two time periods. Indeed, the SA/V in stationary phase never reaches the asymptotic values seen in exponential and transition time periods.

We finish this section on cell dimensions by examining the characteristic that is perhaps most obvious when cells are imaged- their shape, or to be more precise the ratio of their length to their width. In Fig. 4, we plot the fraction of cells that would be perceived as spherical, intermediate, and as rods during each of the periods identified in the growth curve of Fig. 2. We find that the distribution of shapes changes between the periods. In exponential phase, we see the highest fraction of spherical cells for any period at 20% and also the highest fraction of intermediate cells at 43%, whereas rods comprise only 37%. This is the quantification of the visual perception that Fig. 1A is dominated by round and short fat rod cells. During the transition, the fraction of spherical cells is less than a third, and the fraction of intermediate cells is less than half of what they were earlier. Instead rods are by far the dominant shape comprising 74% of the population. This dominance of rods in transition is clearly evident in the visual perception of Fig. 1C. Finally, in stationary phase, the distribution of shapes rebounds back toward the exponential distribution, in that the fraction of spherical and intermediate cells increases at the expense of the rods. However, rods are the dominant shape during this period at 56%, as seen in Fig. 1D and 1E. The morphological changes quantified in this figure explain why previous descriptions of A. baylyi vary in the literature (4, 8).

## Cell surface topography

After studying thousands of cells in images such as those shown in Fig. 1, we identified five distinctive cell surface morphologies we designate as bite, divot, canoe, smooth and filament. These types are shown in Fig. 5A. All of the close-ups in Fig. 5 are taken from the larger images presented in Fig. 1 so that their context can be observed. For each cell type we created a stringent quantitative definition of its morphology as detailed in the Methods. We used these definitions to classify every cell measured. Brief descriptions of the cell types are as follows.

Bite cells are characterized by a single deep, relatively steep-sided indentation that occurs midway along the cell’s long axis. This indentation can be bilaterally symmetric or not- an example of each is shown. Divot cells are characterized by a pair of small depressions occurring near the cell’s poles. A cross section of a divot cell appears in Fig. 1F (lower graph) with the divots marked by arrows. Canoe cells have central surface depression circumscribed by a higher, continuous ridge along the cell edge. A cross-section of a canoe cell appears in Fig. 1F (upper graph) with the center of the canoe marked by an arrow. Smooth cells are featureless with minimal elevation changes on their surface. Filaments are all cells with length greater than 4 µm regardless of any surface features. Filaments may have surface features like those previously described, as in this case a canoe, but if sufficiently long will be classified as a filament.

In addition to classifying cells into the morphological categories shown in Fig. 5A, we also observe cells that have clear septa at their midlines as shown in Fig. 5B. As detailed in the Methods, a septating cell is defined as one whose width at the septum is 50%-90% of its full width. Surface features on septating cells indicate these cells are septating, rather than two cells of that type. For example, in the septating canoe cell shown, the high continuous ridge along the cell edge dips at the septum, and in the septating divot cell there are divots only at the extreme poles of the cell. For all cells measured, we noted whether such septa were present, and created a separate overlaid classification of cells as septating.

Using the strict definitions detailed in the methods, 62% of the total cells measured fall into one of the five cell types shown in Fig. 5A. The remaining cells had surface features that were a combination of one of the types shown or were simply too irregular to classify using our strict definitions. Many published AFM images of bacterial cells show the kinds of surface features seen in Fig. 5 (14, 16–19), including A. baumannii (20, 21).

Figure 6. Relative frequency of occurrence for cell types over a growth curve. The upper graph shows the relative frequency of occurrence for the five cell types identified in Fig. 5A during exponential phase (hours 2–4 after inoculation), the transition (hours 5–7), and stationary phase (hours 8–12). The lower panel gives the percentage of cells observed to be septating during those same three time periods.

In Fig. 7, we look at the average length, width and SA/V of the cell types over the growth curve. Focusing first on the horizontal axis of the figure, we look at length variation. In exponential phase, smooth and bite cells have much same average length, whereas canoe cells are shorter. Divot cells are longer than either of the three types and septating cells are the longest. During the transition, all the cells have shifted being longer, and each cell type is longer than its respective type was exponential phase. Nonetheless, the relative size relationships among the types still holds. Smooth, bites and canoes are the shortest cells, divots are longer, and septating cells are the longest of all. In stationary phase, cells shift back to shorter lengths, and again all cell types are shorter than their respective types during the transition. For the third time, the same trends hold regarding the relative lengths of the cell types.

We now focus on the vertical axis of Fig. 7 to examine trends in the width of cell types. In exponential phase, canoes and smooth cells have similar widths with bites a bit narrower, whereas divots and septating cells are wider. During the transition from exponential phase to stationary phase, all the cell types have narrower widths. Bites and smooth cells have similar widths, again with divots and septating cells wider. However, now the canoes are very narrow as compared to all other cells during this time. In stationary phase, all the cell types have again shifted to lower widths. A similar pattern holds regarding the relationships between cell types- bites, smooth and canoes are the narrowest cells, with divots and septating cells wider.

On Fig. 7 we also plot SA/V ratio as the colored background behind the graph, and the lines show curves of constant SA/V for a cell assuming it is a rod with a hemispherical end cap. For this geometry, the SA/V will increase if a cell becomes either shorter or narrower. From this we observe that cells in exponential phase, which have about the same lengths as those in stationary phase, have nonetheless much lower SA/V because they are wider. We take an example of canoes which have almost the same length in the two phases. Nonetheless, in exponential phase the average SA/V is 4.63 µm− 1. whereas in stationary phase the SA/V is 5.78 µm− 1, an increase by a factor of 25% which is almost entirely mediated by changes in width.

## Cell division mutant

In reading the literature we were struck by the resemblance between our bite cells and cell division mutant Escherichia coli cells (22, 23). To explore this resemblance, we imaged minC mutants of A. baylyi during exponential phase. The results are shown in Fig. 8.

Figure 8A shows that the minC mutant consists mostly of filamented cells, as expected because minicells likely wash away during sample preparation. Along the length of these filaments are indentations that can be compared to those observed in wild type cells shown in Fig. 1B. Zooming in on the boxed cell in Fig. 8A, in Fig. 8B we show a close-up of the one of the indentations, which can be compared to the close-up bite cells in Fig. 5A.

## Cellular morphology in varying culture media

Changes in cellular morphology during different phases of a typical growth curve in general purpose media were presented in Figs. 1–7. In this section we present morphology data from exponential phase cells grown in varying culture media. The data presented in this section will allow us to draw comparisons between cells in various growth stages and cells growing in media of varying nutritional value.

To this end, we grew cells in each of the following media: lysogeny broth (LB), lysogeny broth supplemented with ethanol (LE) or glucose (LG), or minimal media with ethanol (ME), glucose (MG), or 2–3 butanediol (MB) as carbon sources. These six media were selected to attempt to provide the most variable growth rates achievable in < 24 hours of growth in a flask. 2–3 butanediol is also of historical interest because it was used in the enrichment that led to the isolation of A. baylyi strain BD4 (6).

For each medium, we made samples of exponential phase cells from four biological replicates, and we used AFM to image each sample. To quantify the images, we measured the length and width of ~ 75 cells in each of the four images for each of the six media, for a total of 1900 cells assessed. In addition, we took growth curves to find the doubling time for cells in that medium. These data are used to generate the graphs shown in Fig. 9.

From Fig. 9, we make the following observations. The doubling time for all cells grown in LB independent of the additive is about the same. In addition, for all cells grown in LB the length (Panel A), width (Panel B), and SA/V ratios (Panel C) do not substantively differ with differing additive. A comparison between Fig. 9 and Fig. 2 shows that the average length, width and SA/V of cells grown in LB + any additive is in the same range as exponential phase cells in Fig. 2, which were grown in LB.

For the cells grown in minimal media the story is quite different. In this case, the doubling time varies substantially with the additive. As seen in Fig. 9, cells with longer doubling times are substantially shorter (Panel A) and narrower (Panel B), which means they have consequently higher SA/V ratios (Panel C). We obtain an estimate for how size changes with doubling time by fitting the three minimal media points in each of the panels of Fig. 9. From this we find that length decreases at a rate of 0.47 µm/doubling time hour, the width decreases at a rate of 0.12 µm/hour, and the SA/V increases at a rate of 1.06 µm− 1/hour.

In Fig. 10 we show a typical image of cells having the highest doubling time and smallest average size, those grown in minimal media with added glucose. A visual comparison between Fig. 10 and Fig. 1A (both 15 µm images) yields a visceral impression for the size difference.

In addition to changing cell size, changing the additive in the growth medium changes cell shape as shown in Fig. 11. Cells grown in ME have a similar shape distribution to cells grown in LB during exponential phase (compare to Fig. 4). For the cells with increasing doubling times, those grown in MB and MG, the fraction of non-rod shapes increases steadily, so that for MG almost all cells are spherical or intermediate. Clearly Acinetobacter baylyi cells alter their aspect ratios depending on the broth used for growth, which could account for some of the variation in reports of their shape in the literature (6, 7, 24, 25).