When eyes were infected with bacteria containing the pUCP18 plasmid, the duration of the experiment was limited to 2 days to ensure plasmid retention. The plasmid was retained by the bacteria until the end of the 2-day observation period. This was shown by comparison of the number of colonies that grew when eye homogenates were plated on agar with and without carbenicillin. As expected, wild-type PA103 was more virulent than the
exoUexoT double mutant that was complemented with the empty plasmid vector (
P = 0.0017;
Table 5 ). Complementation of the
exoUexoT double mutant with either
exoU or
exoT using that plasmid restored disease severity so that it was no longer significantly different from wild type (for
exoU,
P = 0.8983; for
exoT,
P = 0.1102). The effects of
exoU and
exoT complementation were statistically significant (
P = 0.0181), with significantly worse disease associated with
exoU. Although these results with the overall scoring system showed no statistical difference between disease caused by wild-type and
exoT complemented mutant, it was apparent from the photographs that the disease characteristics differed
(Fig. 2) . This difference was revealed when statistical analysis of total scores, which take into account individual disease parameters, was performed (
P = 0.0073). More detailed analysis showed statistical differences in area of main opacity (
P = 0.0088), surrounding opacity density (
P = 0.0181), and surface regularity (
P = 0.0027), but not in density of main opacity (
P = 0.0845). Because these experiments were terminated after 2 days to minimize plasmid loss, differences in corneal pigment deposition and limbal vessel dilation occurring later in the disease process (7 days after inoculation) were not evaluated.