Sewall and Rodgveller: Changes in body composition and fatty acid profile during embryogenesis of Sebastes maliger 
211 
opment, OGV was plotted against developmental stage 
and the strength of the correlation calculated. The rela- 
tionship between OGV and energetic status of larvae 
was assessed by treating lipid mass, lipid concentration, 
and protein mass as response variables and OGV as a 
predictor variable in simple linear models. Significance 
tests were performed with a one-way analysis of vari- 
ance (ANOVA). 
Fatty acid analysis 
FA composition of total lipid extracts was determined 
by gas chromatography and mass spectrometry. To pre- 
pare lipid extracts for FA analysis, whole lipid extracts 
underwent acid-catalyzed transesterification to fatty 
acid methyl esters (FAMEs), following a procedure out- 
lined by Christie (2003). Two mL of Hilditch reagent 
(0.5 N sulfuric acid [H 2 S0 4 ] in methanol) was added 
to an aliquot of lipid extract which contained 0.3 mg of 
lipid. Before transesterification, 2050 nanograms (ng) 
of 19:0 FA in 50.0 pL hexane was added to each sample 
as an internal standard for quantification. The solution 
was incubated at 55°C for approximately 18 hours, and 
then washed with 5 mL of 5% aqueous sodium chlo- 
ride (NaCl). To separate and extract the FAMEs from 
the aqueous solution, 4 mL of hexane was added, the 
solution was stirred on a vortex mixer, and the hexane 
layer transferred by pipette to a second container; this 
process was repeated with another 4 mL of hexane. 
Four milliliters of 2% potassium bicarbonate (KHC0 3 ) 
was added to the hexane containing the FAMEs to 
quench the esterification reaction and neutralize any 
remaining acid. The hexane-FAME layer was run 
through a sodium sulfate (Na 2 S0 4 ) drying column 
to remove any residual co-extractables and water, 
and the resulting hexane-FAME volume reduced to 
approximately 1 mL in a Labconco Rapidvap (Labconco 
Corporation, Kansas City, MO). Prior to GC analysis, 
2040 ng of 21:0 FAME in 50.0 pL hexane as an instru- 
mental internal standard was added to each sample for 
use in sample recovery calculations. The FAMEs were 
then eluted with a temperature gradient on a Hewlett 
Packard 6890 gas chromatograph (Hewlett-Packard 
Company, Palo Alto, CA) with a 5973 mass selective 
detector by using a 30-m Omegawax 250 fused silica 
column ( Sigma-Aldrich, St. Louis, MO). Five-point 
calibration curves were created from known concentra- 
tions of a Supelco FAME-37 standard mix (Supelco, 
Bellefonte, PA). Thirty of the 32 FAMEs investigated 
yielded calibration curves with a coefficient of deter- 
mination r 2 > 0.990. As a quality assurance measure, 
selected calibration standards were re-injected and 
quantified, and the average across all FAME analytes 
fell within ±1.5% of the known value. 
Along with the samples, quality control samples from 
the lipid extraction step were subjected to the trans- 
esterification procedure. Concentrations of 23 of the 28 
FAMEs detected in the Standard Reference Material 
1946 were within 25% of the average values obtained 
from six previous analyses, with none exceeding 35% 
error. Duplicate larval samples yielded FA concentra- 
tions with coefficients of variation less than 10% for 
26 of the 29 FAs present. Six FAMEs were detected in 
the method blank (in order of mass: 18:0, 16:0, 22:ln-9, 
17:0, 18:ln-9 cis and trans, and 14:0) and the masses of 
these were subtracted from the masses of those FAMEs 
found in each of the samples as a correction. 
Statistical analysis 
To determine whether FA profiles of early-stage embryos 
differed from hatched, preparturition larvae, raw data 
on FA concentrations (ng of FA per g of wet sample 
mass) were first converted to proportions of total FAs 
per sample. The relative proportions of individual FAs 
present in four samples of early-stage embryos were then 
compared to those found in four late-stage samples by 
analysis of similarities (ANOSIM), a nonparametric, 
multivariate statistical test suitable for compositional 
studies (Clarke and Warwick, 1994). ANOSIM was per- 
formed on a dissimilarity matrix based on the Aitchison 
distance (Aitchison, 1992) between all possible pairs 
of samples. The Aitchison distance ( D Aitchison ) between 
two samples, A and B, is derived from the differences 
between the log ratios of pairs of FAs present in the 
two groups: 
D 
Aitchison 
(A.B) 
KJ 
A 
log 
A 
Bj) 
( 1 ) 
where j takes on values up to the number of analytes 
investigated — in this case, 32. 
The Aitchison distance cannot be calculated in cases 
where the concentrations of a FA are zero in any of 
the samples being compared. This proved not to be a 
significant limitation because only one FA, 18:ln-ll, 
was present in measurable quantities in some samples 
but not in others. This FA was excluded from ANOSIM 
analysis, but included in estimates of FA mass losses. 
Three other FAs (15:ln-5, 17:ln-7, and 18:2n-6 trans ) 
were also excluded from analysis because they yielded 
zero values for all samples. We used ANOSIM to com- 
pare the ranked Aitchison distances among samples 
within groups and among samples between groups. This 
yielded the ANOSIM R statistic, which can range in 
value from -1 to 1, with a zero value indicating identi- 
cal groups (i.e., all FAs were used at the same rate, 
resulting in no difference between FA compositions of 
embryos and hatched larvae), positive values indicating 
dissimilarity between groups (i.e., FAs were used at 
different rates, resulting in changes to the FA compo- 
sitions of embryos as they developed into larvae), and 
negative values indicating greater dissimilarity within 
than between groups (i.e., a study design problem). The 
significance value was determined through permuta- 
tions where the observed R value is compared to simu- 
lated R values assuming no difference between groups 
