1.1.

SAR-OCT

Although scattering angle resolved optical coherence tomography (SAR-OCT) instrumentation utilized in this study is described in detail by Gardner et al.,¹⁵ a brief overview is given here. The SAR-OCT instrumentation uses a narrow linewidth tunable laser source and is simple modification of standard swept-source OCT. As with conventional OCT, the sample path of the SAR-OCT interferometer uses an afocal scanning system with the pupil conjugate to lateral scanning mirrors. The SAR-OCT system utilized in this study uses a swept-source laser (1310 nm, $\pm 70\mathrm{nm}$; 100 kHz sweep rate) from Axsun, Inc. (Billerica, Massachusetts, United States). The system has a lateral resolution of $24.0\mu \mathrm{m}$ and an axial resolution of $11.7\mu \mathrm{m}$ in air.

SAR-OCT differs from conventional OCT through the positioning of a transparent PME into the sample path of the interferometer at a plane conjugate to the pupil of the afocal scanning system. The PME utilized in this study is radially angle-diverse and separates incident and backscattered light at lower and higher angles into four different pathlengths. The four pathlengths introduced by the PME correspond to hh, hl, lh, and ll, where $h$ is high-angle and $l$ is low-angle. Because hl and lh are degenerate or equal pathlengths, SAR-OCT instrumentation records the B-Scan sub-images that we label here as $H$ (hh), $M$ (hl and lh), and $L$ (ll). Using a Fourier optics approach, Yin et al.¹⁶ derived analytical expressions for the SAR-OCT optical transfer functions for $H$ (hh); $M$ (hl and lh); and $L$ (ll). The SAR-OCT transfer functions [Eq. (23) $H_{l,l}$, Eq. (24) $H_{h,h}$, and Eq. (25) ${\mathrm{H}}_{h,l}$ and $H_{l,h}$ in Yin et al.] are the Fourier transforms of the point-spread function and give the complex amplitude ratio between spatial-oscillations in object and image planes. An important element in Yin et al.’s analysis is the sample backscattering angular diversity function [Eq. (6) in Yin et al.]. By assuming the backscattering angular diversity function has a Gaussian form, Yin et al. showed that the SAR-OCT L/M intensity ratio decreases with increased angular variance [$\sigma$ in Yin et al., Eq. (27)] of backscattered light. Larger angular variance ($\sigma$) gives smaller SAR-OCT L/M ratios, whereas more narrow angular backscattering variance (i.e., small $\sigma$) give larger L/M intensity ratios.

2.1.

Bacillus subtilis Colonies

Bacillus subtilis was selected due to similar size to mitochondria (1 to 10 microns) and with analogous rod-to-coccus morphological transitions depending on temperature. Average thickness of bacterial colonies studied were 480 microns. Bacillus was cultured in five different Petri dishes to produce five distinct colonies. Prior to imaging, each dish was sampled/swabbed twice to create two sister colonies in two other dishes, which were used for gram staining and a control. Thus, a total of 15 Petri dishes were prepared (Fig. 1).

Fig. 1

Bacterial imaging and processing. Each Petri dish represents three trials. Control plates were never exposed to SAR-OCT. The SAR-OCT control plates were never exposed to heat. Heating of the experimental plates were from $t=0$ throughout SAR-OCT imaging.

2.2.

Bacillus Subtilis Imaging and Data Collection

Five imaging trials were conducted. In each trial, an experimental, SAR-OCT control, and gram stain control Petri dishes were imaged. Each of the dishes within a trial had Bacillus sister colonies. The first Petri dish was not subjected to temperature change and was gram stained and imaged using a light microscope to set a baseline for initial Bacillus morphology. The second dish was imaged by SAR-OCT while gradually being heated from room temperature (21°C) to 42°C over at least 100 min to induce a rod-to-coccus morphological change.^13,14 The induction of rod-to-coccal morphology has been previously described and the time point selected allowed for sample-wide morphological changes without inducing cell wall thickening secondary to cross linking.¹⁵ Once SAR-OCT imaging was complete, the bacterial colonies were gram stained again and imaged via light microscopy to validate a transition from rod-to-coccus morphology occurred. The third Petri dish served as an imaging control and was subjected to the same temperature change as the second Petri dish and underwent gram staining but was not imaged via SAR-OCT. The third Petri dish was heated to 42°C while in contact with a sand bath. Sand was utilized to minimize movement of the Petri dish during the duration of the experiment and ensure consistent and repeatable SAR-OCT imaging throughout the experiment. Temperature of the bacterial colony surface was recorded using a noncontact infrared thermometer. Once the infrared thermometer indicated 42°C, a 105-min timer was initiated, as detailed in Burdett et al.’s imaging protocol.¹⁵

2.3.

Analysis SAR-OCT Images of Bacillus subtilis Colonies

Image segmentation and fitting the L/M distributions was applied to SAR-OCT data to detect the rod-to-coccus transition. From SAR-OCT data, raw images were compiled into 512 ($x$) by 512 ($y$) by 160 ($z$) voxel $B$-scan stacks, which were large enough to capture individual colonies. Several colonies were captured in each Petri, providing at least 30 colony samples per trial. To separate the bacterial images from the Petri dish, Otsu’s thresholding was applied to SAR-OCT images of Bacillus colonies. After sampling only pixels relating to the Bacillus, images were flattened and the L/M ratio was calculated over every pixel in the Bacillus samples. Bacterial colonies were analyzed over five trials, incorporating hundreds of cells that, while directly analyzed in gram stain, are analyzed as whole colonies during SAR-OCT intensity analysis.

After segmentation, L/M intensity ratios for cocci were compared to values before the rod-to-coccus transition. To further differentiate between rod and coccus morphologies, a histogram of L/M ratios was compiled by considering active pixels within a Bacillus colony.

2.4.

Analysis of Gram Stain Images of Bacillus subtilis Colonies

Changes in average bacterial length of a colony in images of the gram stains were used to validate a rod-to-coccus transition occurred. Bacterial length was manually extrapolated based on overall image scale. Length was calculated for all bacteria within a randomly selected quadrant of the gram stain image (Fig. 3).

2.5.

Analysis of Bacillus Orientation

The three-dimensional architecture of colonies of rod-shaped bacterial (Vibrio Cholera, E. Coli, Bacillus subtilis) has been studied from both theoretical and microscopic perspectives. Bacteria placed on agar form a monolayer biofilm initially with the long axis of the bacteria parallel to the surface of the agar plate. As the biofilm grows, the bacteria begin to stack in the center with their long axis perpendicular to the agar plate. Eventually the colony assumes a three dimensional hill-like structure with bacteria at the surface oriented parallel to the agar plate and bacteria in the depth of the center oriented perpendicular to the agar plate. Studies have demonstrated, using confocal imaging and computational modeling, that bacterial orientation in a colony varies from surface to bottom and from center to periphery. Thus, distribution of L/M ratios are expected to vary with location in a colony. Literature shows that Bacillus at the center and toward the bottom and middle of the colony by depth will be oriented with the long axis of the bacteria largely perpendicular to the plane of the agar plate.^17–22 Alternatively, Bacillus on the colony surface and towards the peripheral edges will be oriented with their long axis parallel to the surface agar in the plate.

Based on basic scattering theory and Yin et al.’s results, these documented variations in cell orientation are expected to impact SAR-OCT L/M ratios at different regions in the Bacillus colonies. To investigate whether SAR-OCT can detect changes in bacterial orientation documented by Warren et al.,¹⁷ Panning windows ($15\times 15$) were applied to SAR-OCT B-scan vertical slices of the colonies. L/M ratios were sampled from and averaged within the panning window. Panning windows were applied to every pixel in the colony images.

3.1.

Bacillus subtilis Whole Colony Gram Stain Analysis

Microscopy of bacterial colonies confirmed consistent controls and initial conditions across all five trials. Bacterial colonies demonstrate a rod-shaped morphology in dish one in all five trials and no significant difference in average bacterial length or bacterial clustering density between dishes two and three over all five trials. Gram stains conducted on bacterial colonies confirm the predicted rod-to-coccus transition in bacterial morphology after heating. On average, bacterial length before heating was $7.42\pm 0.6\text{microns}$ and $4.8\pm 0.8\text{microns}$ after heating ($p=0.038$), and bacterial orientation was random on the gram stains (Fig. 2).

Fig. 2

Bacterial gram stains: (a) image of a gram stain of colonies with no heating and (b) image of a gram stain of colonies after heating. Average bacterial length before and after heating were $7.42\pm 0.6\text{microns}$ and $4.8\pm 0.8\text{microns}$, respectively ($p=0.038$). The scale bar in both images is $50\mu \mathrm{m}$.

3.2.

Bacillus subtilis Whole Colony L/M Ratios

A significant change in L/M intensity ratio was observed preheating, 1.48 (mean) ± 0.608 (standard deviation) versus post heating, $1.2086\pm 0.6679$ (s) with a $T$-test $p=0.032$. When a composite L/M ratio image was formed and its intensity histogram was fit to the Burr distribution, the $C$-parameter (representing approximate width of the distribution could be applied to discern between rod and coccus Bacillus: average $C$-parameter for preheating is $4.1\pm 0.33$ versus postheating is $5.4\pm 0.27$ ($p=0.047$). While individual L/M ratios alone could not significantly distinguish between rod and coccus bacterial colonies, $C$-parameters derived from Burr distributions provided statistically significant and robust discrimination.

3.3.

Bacillus subtilis Orientation Analysis

Four factors are recognized where differences in Bacillus orientation may affect the SAR-OCT L/M ratio. The conditions are illustrated in Fig. 3.

Fig. 3

SAR-OCT L/M intensity ratio and relationship with Bacillus orientation.

Based on basic scattering theory and Yin et al.’s results, SAR-OCT L/M intensity increases with the lateral spatial extent of the scatterer with respect to the incident beam. Thus, Bacillus in different locations in the colony and with different morphologies after heating will have different L/M distributions.

Panning window sampling and averaging of the L/M ratio across Bacillus colony $B$-scans demonstrated that as expected the L/M ratio varied based on location within the colony—surface versus bottom and center versus periphery. L/M varied from 1.44, at the colony surface, to 1.03 at the bottom and center of the colony. Considering that long axis of bacteria at the colony surface are predicted to be perpendicular to the direction of incoming OCT light and the bacteria in the bottom are predicted to be parallel to the image beam, L/M values were translated to angles based on an empirically derived equation: (L/M-1.03)/0.41 = $\cos (\mathrm{\Theta })$. Here, the offset value 1.03 represents the minimum L/M value in the experiment and 0.41 is the observed variation of L/M values and is used to scale angles ranging from 0 deg to 90 deg relative to the plate surface. Using this empirical equation, several images (Fig. 4) were computed indicating Bacillus orientation.

Fig. 4

SAR-OCT $B$-scan-derived images of Bacillus orientation in a colony. L/M ratio varies from 1.03 to 1.44 across all bacterial colonies. Assuming 1.03 (L/M) corresponds to a 90 deg orientation and 1.44 to be 0 deg from the agar plate surface, respectively, the arrow window plots were overlaid on the bacterial colony to indicate predicted Bacillus orientation. An empirically derived equation gives Bacillus orientation (Θ) under these imaging conditions: (L/M−1.03)/0.41 = $\cos (\mathrm{\Theta })$. Panels (a)–(c) are B-scans from different Bacillus colonies, with (a) and (b) having the raw B-scan coupled with L/M arrow window plots. L/M ratios are highest at the surface of the colonies.

3.4.

Variance in L/M Ratios after Heating by Colony Location

Results of L/M change following heating of Bacillus colonies also followed qualitative predictions from basic scattering theory and Yin et al.’s results and are summarized in Table 1.

Table 1

L/M ratios postheating. Results follow predicted increases and decreases in L/M ratio for Bacillus at the bottom and center regions of a colony and Bacillus at the surface and periphery of the colony, respectively. All changes were statistically significant p<0.05.