All eyeballs were tissue paper blotted to enable bacterial adherence to the corneal epithelium. Susceptibility to traversal by adherent bacteria was then toggled on/off by use/omission of EGTA treatment following the blotting procedure. Upper images (EGTA used to toggle on traversal) show deep bacterial traversal 6 h after bacterial challenge in EGTA-treated corneas. Lower images were PBS-treated controls (EGTA step omitted) and showed bacterial adherence to the surface without subsequent bacterial traversal. Panel (a) shows confocal technique using mice with CFP-tagged cell membranes (cyan), and panel (b) shows two-photon images of cellular NAD(P)H autofluorescence (magenta) in transgenic (C57BL/6 background) or wild-type (Black Swiss background) mouse corneas respectively. Field size: 189 µm × 189 µm.

Panel (a) shows NAD(P)H autofluorescence (magenta) of corneal epithelial cells and bacteria (green) detected from optically opaque (infected) mouse eyes 24 h post-scratch and inoculation. In panel (b), bacteria (green) within the stroma of the cornea were found to orient themselves along the pattern of the collagen fibrils. Field size: 189 µm × 189 µm.

Panel (a) shows two-photon images of cellular NAD(P)H autofluorescence (magenta) of MyD88 (−/−) (top) and wild-type (bottom) mouse corneas. Deep bacterial traversal (GFP, green) of corneal epithelium was detected in MyD88 (−/−) but not wild-type mice 8 h after bacterial challenge ex vivo. Only adherent bacteria without epithelial traversal were found in tissue paper-blotted wild-type corneas. Panel (b) shows an earlier (4 h) time-point at which a relatively larger number of bacteria were found adhering to the MyD88 (−/−) corneal epithelium as compared to the wild-type. In panel (c), merged image (fuchsia) of reflection (red) and autofluorescence (magenta) is shown. Reflection confocal methodology (red) can be used to capture cells not visualized by autofluorescence (i.e. dying or dead cells) to monitor cell viability during bacterial traversal. This method confirmed that penetrated bacteria at the later (8 h) time point were overlaid with epithelial cells.

Fluorescein was added to tissue paper blotted or normal healthy (non-blotted) mouse corneas to access cell-cell junction integrity. Extent of corneal staining was examined using a slit lamp (a-c) and confocal microscopy (d–i). Panels (a, d, g) show extensive fluorescein staining (green) in blotted wild-type (C57BL/6) corneas, but not non-blotted wild-type (b, e, h) or MyD88 (−/−) (c, f, i) mouse corneas. Z-stack confocal images (1272 µm × 1272 µm × 100 µm) are presented as 3D block view (d, top), 3D orthoslice view (d, bottom), and 2D x-y view (g) for blotted and fluorescein stained wild-type mouse cornea. In parallel, 3D block views and 2D x-y views of confocal images are shown for normal wild-type (e, h) and MyD88 (−/−) (f, i) mouse corneas. These data show that epithelial tight junctions are intact in MyD88 knockout corneas prior to bacterial exposure.

Three different techniques were used to visualize cells in all layers of the cornea. Panels (a) and (d) show two-photon excitation of cytoplasmic NAD(P)H to reveal metabolically active (live) cells in the epithelium and endothelium of the cornea by autofluorescence (magenta). Panels (b) and (e) show reflection confocal imaging (red) for localization of epithelial and endothelial cells (including live and dead cells), and collagen fibers in the stromal region (red). Panels (c) and (f) show confocal imaging of CFP-tagged cell membranes (cyan) expressed in transgenic mice. All three imaging methods enabled visualization of distinct cell layers in intact corneas of at least 150 µm thickness (i.e. mouse). Panel (g) shows two-photon autofluorescence and reflection images overlaid to distinguish live and dead cells. Panel (h) shows that 3D images of CFP-membranes exhibit a high signal-to-background ratio providing excellent contrast for bacterial localization experiments. Field (xy) size: 189 µm x 189 µm.

Traversal was enabled in the corneal epithelium of transgenic mice expressing CFP-tagged cell membranes (cyan). Panels (a, b) show bacterial-induced membrane bleb formation ex vivo visualized as spherical membrane projections (arrow) extending away from the epithelial cells. A representative view in the xz plane is shown in (a), and a higher magnification imaging revealing a bleb-confined bacterium is shown in (b). In (c), bacteria can be seen located between cells (orange arrows) where some were motile (fast-moving GFP-bacteria, red; slower-moving GFP-bacteria, white  =  captured twice in both CFP and GFP channels). Other bacteria in this image appeared to be in the cytoplasm (red arrow).

Panel (a) shows a schematic of the custom bacterial quantification program. Each image stack is divided into ∼1000 sub-volumes; in each sub-volume, the top and bottom of epithelium were localized in which 10 bins along the z-depth were assigned, followed by bacterial fluorescence quantification in each bin, and finally summation of all bins of the same z-depth across the entire image stack. Panel (b) shows quantitative analysis of bacteria traversing to the basal lamina for MyD88 (−/−) but not wild-type mice (0.00 represents the corneal surface, 1.00 the basal lamina). Average ± S.E. of three fields are shown.

Panels (a) and (b) show two-photon images of cellular NAD(P)H autofluorescence (magenta) in non-blotted MyD88 (−/−) or wild-type mouse corneal epithelium 8 h after bacterial challenge ex vivo. Significant bacterial adherence and traversal (GFP, green) were found in MyD88 (−/−) mouse (a) but not wild-type mouse corneas (b). This observation was confirmed by quantitative localization of GFP bacteria using the custom program as shown in panel (c).

Similar to the ex vivo results at the same time-point, panels (a) and (b) show blotted corneas of MyD88 (−/−) mice were more susceptible to both bacterial adherence and traversal after 8 h of bacterial challenge as compared to blotted wild-type which showed bacterial adherence only. This observation was confirmed by quantitative localization of GFP bacteria using the custom program as shown in panel (c). Whereas in the absence of blotting, panels (d) and (e) show bacterial traversal in vivo was disabled. Only bacteria adherence was detected on MyD88 (−/−) (d) but not wild-type (e) mouse corneas. This observation was confirmed by quantitative localization of GFP bacteria using the custom program as shown in panel (f).