Lewis et al : Integrating DNA barcoding of fish eggs into ichthyoplankton programs 
155 
during NEFSC surveys in the winter ( January-Febru- 
ary), late spring (May-June), late summer (August) 
and late autumn (November-December) over a 10-year 
period (2002-2012). Temperature and salinity profiles 
through the water column were collected with an SBE 
19plus V2 SeaCAT 3 conductivity, temperature, and 
depth (CTD) profiler (Sea-Bird Electronics Inc., Bel- 
levue, WA) attached to the tow wire above the bongo 
nets. Sea-surface temperatures (SSTs) were recorded 
during the upcast of the CTD profiler at the shallow- 
est depth bin. Details on the collection localities and 
sampling dates were deposited within the publically 
available BOLD project file entitled “NIFEB: Fish Eggs 
Barcoding.” 
Ichthyoplankton samples preserved in EtOH were 
sorted manually for all fish eggs and larvae in the 
laboratory. Fish eggs were removed from a sample and 
transferred to 7-mL (2-dram) glass vials filled with 
95% EtOH. No morphological identifications were at- 
tempted during or after the sorting procedure. 
Subsampling of eggs for molecular identification 
The total number of eggs collected in all of the sam- 
ples exceeded our molecular processing capacity. For 
that reason, we implemented a 2-stage subsampling 
procedure designed to determine the diversity of eggs 
within each sample. For the first stage of subsampling, 
a maximum number of 10 eggs were randomly selected 
for identification from each sorted sample. Individual 
eggs were first digitally photographed and measured 
(in millimeters) with Nikon imaging software (NIS-E1- 
ements BR, vers. 2.3, Nikon Instruments Inc., Melville, 
NY) and a color digital camera (Nikon DXM-1200C) 
mounted on a stereo microscope (Nikon SMZ1500) 
under both reflected and transmitted light. Fish eggs 
were then placed in 96-well plates with one egg and 
one drop of 95% EtOH per well. A negative control well 
also was included on each plate. 
Of the initial 456 samples, 73 contained >10 eggs 
and were subjected to a second round of subsampling 
that that was based on both egg measurements and 
the results from the initial round of molecular iden- 
tifications. For this second round of subsampling, we 
measured the diameter of the remaining unidenti- 
fied eggs in the samples. Histograms of egg diameters 
within 0.05-mm-diameter bins were developed for each 
sample, and additional eggs were chosen for molecular 
identification from any 0.05-mm-diameter bin that ei- 
ther did not include an egg identified during the first 
round of molecular identification or contained a high 
number of eggs and multiple species of eggs. The intent 
of this second round of subsampling was to ensure that 
the diversity of eggs within a sample was identified, 
while avoiding repeatedly sequencing the same species 
of egg at the same sampling station. As a result of the 
3 Mention of trade names or commercial companies is for iden- 
tification purposes only and does not imply endorsement by 
the National Marine Fisheries Service, NOAA. 
opaqueness of ethanol-preserved eggs, we did not use 
any other morphological feature (other than egg diam- 
eter) for subsampling. 
To speed up the subsampling process, we developed 
an automated egg measuring graphical user interface 
with the Image Processing Toolbox in MATLAB, vers. 
R2012A (The MathWorks Inc., Natick, MA). Up to 40 
eggs at a time were digitally photographed on an acryl- 
ic plate containing a rectangular well specifically sized 
for the digital image taken at 3x magnification (Fig. 
1A). A black and white threshold was then manually 
applied to each image (Fig. IB) and then the Hough 
transformation was applied to the thresholded image 
(Fig. 1C). The Hough transformation is designed to 
find circles within an image, including circles with a 
broken outer border. The interface that we developed 
contained slider bars that allow a user to optimize the 
black and white thresholding process for an individual 
picture, as well as the minimum circle quality used 
in the find circle algorithm. Finally, manual editing 
of the automatically measured circular egg was per- 
formed, and measurements of ovoid eggs were added 
(Fig. ID). The result of this procedure was a text file 
that contained the diameters of each measured circular 
egg or the long axis and short axis of ovoid eggs. 
Egg abundances are reported as the number of eggs 
per 10 m 2 of water. To account for the subsampling, 
each station’s total egg abundance within each 0.05- 
mm egg-diameter bin was first calculated. For each 
diameter bin, species were assigned in proportion to 
the available molecular identifications within that bin 
with a procedure analogous to the use of age-length 
keys. Oval eggs, which corresponded to 1 of 3 anchovy 
species, were excluded from this process because they 
could be assigned to species based solely on shape (long 
axis:short axis) and size. 
Molecular identification protocol 
Prepared plates were sent by mail to the University of 
Guelph’s Canadian Centre for DNA Barcoding, a mo- 
lecular identification laboratory with a proven success 
rate and developed database capable of accommodat- 
ing large data output associated with high-throughput 
DNA sequencing. Key sample data, including specimen 
collection information, voucher image files, and a plate 
record (sample array details for each plate) were sent 
electronically to the Canadian Centre for DNA Barcod- 
ing. This information was organized within the BOLD 
online database, and each COI barcode sequence was 
connected to its source specimen following DNA bar- 
code analysis. 
Standard DNA barcoding protocols were followed for 
all analytical steps, including DNA extraction, PCR, 
and DNA sequencing. Submitted samples were subject- 
ed to overnight lysis in a lysis buffer with proteinase K 
(Thermo Fisher Scientific Inc., Waltham, MA), followed 
by DNA extraction onto a glass fiber membrane (Pall 
Corp., Port Washington, NY) by using an automated 
protocol (Ivanova et al., 2006). A barcode region of 658 
