Klibansky and Juanes: Effects of preservation on oocytes of three Atlantic fish species 



541 



Image software used those grey values to determine 

 the perimeter of the particles; therefore standardiz- 

 ing the upper limit of the density slice among samples 

 would result in less accurate measurements. Then the 

 "analyze particles" command was run on particles from 

 150 to 1500 pixels in area. The process was also set to 

 include interior holes and to ignore particles touching 

 the perimeter as the oocytes were measured. 



The output of this process produced four columns of 

 data: area, perimeter, major axis length, and minor axis 

 length for each particle. These data were then trans- 

 ferred to a Microsoft Excel spread sheet (Microsoft® 

 Office Excel 2003, Microsoft Corporation, Redmond, 

 WA), where a macro was run to filter out measure- 

 ments of non-oocytes by accepting only particles within 

 a narrow range of roundess values. The macro then 

 calculated the mean oocyte diameter of each sample, 

 as well as other descriptive statistics, and produced a 

 percent frequency histogram for the sample of round- 

 ness-filtered particles. 



Ideally all samples would have been analyzed after 

 the same amount of time that they had been preserved, 

 although it was not practical to do so. Thus time (days) 

 preserved was recorded for each sample and the aver- 

 age time that the sample was preserved was compared 

 between groups where appropriate. 



Statistical analyses 



The preserved weight of a lobe or subsample of ovar- 

 ian material was compared to its fresh weight. Percent 

 change in weight was calculated with Equation 1: 



%A^, = {iWtp^^^,,, - Wtp^^, ) / Wtp^, ) X 100, ( 1 ) 



where ^A^y^ = the percent change between Wtp^^^^^^,^^ 

 and Wtp^^^,; 

 ^^Preseroed ~ preserved weight of a sample of ovarian 

 material; and 

 ^^Fresh ~ fresh weight of that sample. 



A positive ^^vy, indicated an increase in weight due to 

 preservation. 



It was not logistically possible to measure the dia- 

 meters of fresh oocytes; therefore it was necessary to 

 use one of the preservative treatments as a control 

 treatment. The formalin treatment was chosen for this 

 purpose, because it is a standard preservation method 

 and thus was expected to have the most consistent effect 

 on oocyte size. In this experiment, the other four treat- 

 ments (Gilsons, ethanol, freezing, and split-formalin) 

 were considered experimental treatments. Change in 

 mean oocyte diameter due to preservation in the experi- 

 mental treatments was quantified by using Equation 2: 



%A0D = {(ODE.rpenn,en>.I " OZ?Co„r™/ » ' ODc„„,^, ) X 100, ( 2 ) 



where ^^^qd - ^^^ percent difference in mean oocyte 

 diameter between OD£^p„,„^„,„, and 



^^ Control' 



ODg^ -^^^^i^i = the mean oocyte diameter of a sub- 

 sample of an ovary preserved in one of 

 the four experimental treatments; and 

 ^^Coniroi - *h^ mean oocyte diameter of a sub- 

 sample of the same ovary preserved in 

 the formalin treatment. 



A positive '^cAq^ indicates that the mean oocyte diameter 

 of a subsample in the experimental treatment is larger 

 than in the formalin treatment. 



For examining "^c \x, and 'Ic Aqj^, samples were grouped 

 by experimental treatment within species, and then in 

 the case of cod where samples were collected from two 

 regions, the samples were further grouped by region. 

 T-tests were conducted within these groups to test the 

 null hypothesis (//q) that '7c A - for each experimental 

 treatment. Assessment of normality of each group by 

 examining boxplots and histograms indicated that data 

 did not require transformation. Statistical analyses 

 were performed with SAS (version 9.1, SAS Institute 

 Inc., Cary, NO. Each test had n^ + n^-l degrees of 

 freedom (Tables 2 and 3) where n-^ and ng were the 

 total numbers of observations from each group. Where 

 t-tests were conducted in groups, significance (o) levels 

 were adjusted by the sequential Bonferroni procedure 

 (Quinn and Keough, 2002), to minimize family-wise 

 type-I error rate. 



When a significant difference in %Aqjj was found 

 between two groups, time preserved was investigated 

 as a confounding factor, although any major change in 

 oocyte size due to preservation typically happens within 

 one day (Kjesbu et al., 1990). A f-test of time preserved 

 between the groups was conducted in which the ratio 

 of time preserved in experimental treatment to time 

 preserved in control treatment was used as a metric 

 of time preserved, because the calculation for %Aqjj 

 includes OD 



Contro] 



in the denominator (see Eq. 2). If no 

 significant difference was found, we concluded that the 

 difference in ^Aqij was not due to time preserved. If a 

 significant difference was found, a one-tailed <-test of 

 %4oD between these groups was conducted. Assuming 

 that the sign of the slope of the relationship between 

 time preserved and oocyte diameter is constant after 

 the initial changes that occur within days of preser- 

 vation, the group with a higher time preserved value 

 should have exhibited a greater WcAqj^]. Thus the Hq 

 for each one-tailed ^-test was: I mean %4qqI of group 

 preserved for more time < I mean %Aqj)\ of group pre- 

 served for less time. If this analysis implicated time 

 preserved as a likely cause of a significant difference 

 in VcAqu, then the effect of the experimental treatment 

 would be considered confounded. 



In situations where there was a significant differ- 

 ence in %Aq,j and time preserved, but the sign of %Aqi^ 

 was different between the groups, no such test was 

 performed. Instead it was concluded that the difference 

 in %Aqu was not due to time preserved, based on the 

 assumption that the sign of the slope of the relation- 

 ship between time preserved and oocyte diameter is 

 constant. 



