© 2015. Published by The Company of Biologists Ltd | Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
RESEARCH ARTICLE
TECHNIQUES AND RESOURCES
SNPfisher: tools for probing genetic variation in laboratory-reared
zebrafish
ABSTRACT
Single nucleotide polymorphisms (SNPs) are the benchmark
molecular markers for modern genomics. Until recently, relatively
few SNPs were known in the zebrafish genome. The use of nextgeneration sequencing for the positional cloning of zebrafish
mutations has increased the number of known SNP positions
dramatically. Still, the identified SNP variants remain under-utilized,
owing to scant annotation of strain specificity and allele frequency.
To address these limitations, we surveyed SNP variation in three
common laboratory zebrafish strains using whole-genome
sequencing. This survey identified an average of 5.04 million SNPs
per strain compared with the Zv9 reference genome sequence. By
comparing the three strains, 2.7 million variants were found to be
strain specific, whereas the remaining variants were shared among all
(2.3 million) or some of the strains. We also demonstrate the broad
usefulness of our identified variants by validating most in independent
populations of the same laboratory strains. We have made all of the
identified SNPs accessible through ‘SNPfisher’, a searchable online
database (snpfisher.nichd.nih.gov). The SNPfisher website includes
the SNPfisher Variant Reporter tool, which provides the genomic
position, alternate allele read frequency, strain specificity, restriction
enzyme recognition site changes and flanking primers for all SNPs
and Indels in a user-defined gene or region of the zebrafish genome.
The SNPfisher site also contains links to display our SNP data in the
UCSC genome browser. The SNPfisher tools will facilitate the use of
SNP variation in zebrafish research as well as vertebrate genome
evolution.
KEY WORDS: Danio rerio, Zebrafish, Genome, SNP, Variation, Nextgeneration sequencing, Whole-genome sequencing
INTRODUCTION
Next-generation sequencing (NGS) has revolutionized the
acquisition of large amounts of genomic data (Turner et al.,
2009). The ability to sequence entire genomes has spawned
projects like the ‘Thousand Genomes Project’, which aims to
identify variation within human populations to make genotypephenotype correlations possible for complex traits (Abecasis et al.,
2010). In contrast to the human genome, much less is known
1
Section on Vertebrate Development, Program in the Genomics of Differentiation,
National Institute of Child Health and Human Development, National Institutes of
2
Health, Bethesda, MD 20892, USA. Section on Molecular Dysmorphology,
Program in Developmental Endocrinology and Genetics, National Institute of Child
Health and Human Development, National Institutes of Health, Bethesda, MD
3
20892, USA. Department of Cell and Developmental Biology, University of
Pennsylvania School of Medicine, Philadelphia, PA 19104, USA.
*Joint first authors
‡
Author for correspondence ([email protected])
Received 24 October 2014; Accepted 25 February 2015
1542
about the genetic variation of the zebrafish genome, despite its
similar size. The zebrafish (Danio rerio) has emerged as one of
the most popular model organisms for phenotype-based forwardgenetic mutagenesis screens owing to its optical clarity, rapid
development and high fecundity (Patton and Zon, 2001). Unlike
other vertebrate model organisms, most of the common laboratory
zebrafish strains were established relatively recently from single
pairs of founder fish caught in the wild or purchased from pet
stores, with progeny selected for lack of embryonic lethality and
developmental anomalies (Trevarrow and Robison, 2004).
Generally, subsequent propagation of these wild-type strains has
sought to avoid loss of genetic variability due to bottlenecking
and genetic drift by avoiding inbreeding, maintaining a maximum
effective population size and importing individuals of the same
strain from other populations (Harper and Lawrence, 2011).
However, most labs maintain their own small, closed populations
of each wild-type strain, and adoption of the aforementioned
measures varies greatly among fish facilities (D. Castranova,
personal communication). Consequently, the genetic variability
between and within wild-type strains probably varies considerably
among different population isolates.
Only a handful of studies have attempted to quantify the genetic
variability of laboratory zebrafish strains. Empirical measures of
genetic variability in wild-type zebrafish strains have suggested
high levels of both intra- and interstrain variation. Initial studies
using simple sequence length polymorphism markers (SSLPs) were
limited by either the number of markers used or by the number of
individuals of a strain tested (Knapik et al., 1998; Nechiporuk et al.,
1999; Coe et al., 2009). Similarly, subsequent studies using single
nucleotide polymorphisms (SNPs) greatly expanded the number of
markers tested, but few individuals of each strain were genotyped
(Stickney et al., 2002; Guryev et al., 2006; Bradley et al., 2007;
Whiteley et al., 2011). Recently, whole-genome sequencing (WGS)
has been used to identify SNP and insertion/deletion (Indel)
variation in wild-type and chemically mutagenized zebrafish
(Bowen et al., 2012; Obholzer et al., 2012; Howe et al., 2013;
Patowary et al., 2013). Many of the NGS positional cloning
strategies described for zebrafish rely heavily on the ability to filter
out normal variants and/or track strain-specific SNPs (Bowen et al.,
2012; Leshchiner et al., 2012; Obholzer et al., 2012; Voz et al.,
2012; Miller et al., 2013). Therefore, NGS positional cloning
strategies would be enhanced by increased information on normal
genetic variants found within the zebrafish genome. This would
facilitate differentiation between neutral polymorphisms and
phenotype-causing mutations, particularly when no ‘smoking
gun’ mutations are identified in the critical regions.
In this study, WGS was employed to find SNP and Indel variant
positions in three populations of commonly used laboratory zebrafish
strains (EK, TL and WIK). An average of 5.04 million SNPs was
identified in each strain surveyed. Each SNP was categorized
DEVELOPMENT
Matthew G. Butler1,*, James R. Iben2,*, Kurt C. Marsden3, Jonathan A. Epstein2, Michael Granato3 and
Brant M. Weinstein1,‡
according to its strain distribution, frequency and position (exon,
intron, etc.) in the genome. The resulting SNP variation dataset was
then organized into a user-friendly suite of tools named ‘SNPfisher’.
SNP variants were also formatted for display in the UCSC genome
browser for integrated genomic displays (Kent et al., 2002).
RESULTS
WGS of three common laboratory zebrafish
WGS was performed to identify both SNP and Indel variation in
three common laboratory zebrafish strains, including Tüpfel long fin
(TL), WIK and Tg( fli1a-eGFPy1). TL was originally purchased in a
pet shop and derives its name from the recessive leopard phenotype
(tüpfel translates to ‘dot’ in English) and dominant long fin mutant
phenotypes (Tresnak, 1981; Haffter et al., 1996). WIK, formerly
WIK11, originated from a single pair of wild-type zebrafish caught
in India exhibiting the absence of embryonic and larval lethality
(Rauch et al., 1997). The Tg( fli1a-eGFPy1) transgenic strain,
popular in vascular biology and henceforth referred to as FLI, was
originally generated from and has been subsequently back crossed
to the EkkWill (EK) strain [originally derived from fish obtained
from EkkWill Tropical Fish Farm (Gibsonton, FL, USA)] and
therefore represents an isolate of EK (Lawson and Weinstein, 2002).
Sequencing libraries were prepared from genomic DNA isolated
from each of the above strains separately. Each strain library was
generated from caudal fin clips from 18 individual adult zebrafish
(nine females and nine males), pooled into a single sample for
library construction. Illumina HiSeq sequencing reads were aligned
to the Zv9 reference genome sequence, a genome sequence mainly
generated from the Tübingen strain (TÜ) (Howe et al., 2013). An
alternative allele was considered to be anything that differed from
the Zv9 reference sequence at a given position. Evaluation of the
resulting alignments demonstrated a mean of 16.1× genome
coverage for each strain surveyed. With genome coverage of 16×,
we expect to detect 80% of alternate alleles with frequencies equal
to 0.1 assuming random sampling (supplementary material Fig. S1).
However, a minimum read depth of three reads was required for
inclusion in our dataset, which drops the expected detection to 78%
for alternate alleles with frequencies equal to 0.4 (supplementary
material Fig. S1). Prior to applying filters, Genome Annotation
Toolkit (GATK) marked 35.4±0.2 million reads as containing a
variant per strain (supplementary material Fig. S2). The total
number of variant reads was reduced to 10.7±0.2 million (∼30% of
total) once quality control filters (described in the Materials and
Methods) were applied (supplementary material Fig. S2). The
remaining variants were then further filtered by requiring a 3-read
minimum in each strain and by removing any variants contained in
repetitive sequence as annotated by Repeatmasker. After this
filtering, the surveyed strains averaged 5.04±0.22 million variants
per strain. This equates to a SNP ratio of seven SNPs per 1 kb of
unique sequence in the zebrafish reference genome. In order to put
the amount of variant positions in laboratory zebrafish into
perspective, a scripted calculation was used to determine the total
number of SNP variants in a single, ethnically defined human
population when 18 individuals were randomly chosen from the
1000 genomes samples (Abecasis et al., 2012). Surprisingly,
laboratory zebrafish ranged from 6.3 to 4.4 SNPs per 1 kb, which is
a higher ratio than any ethnically defined human population (2.6-4.4
SNPs per 1 kb) measured (supplementary material Fig. S3). This
suggests that either the human genome reference sequence is more
representative of the consensus human genome among various
populations or there is more genetic variation in the wild zebrafish
populations that were used to establish laboratory strains.
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
Strain-specific and shared genetic variation in laboratory
zebrafish
For many applications, especially genetic mapping, the strain
specificity and allele frequency of a polymorphism will determine
its utility. Therefore, SNPs and Indels that passed each of the filter
criteria were classified according to their distribution among each
strain (Fig. 1A). For purposes of our analysis, strain-specific variants
were defined as alternate alleles observed in only one of the strains.
Strain-specific SNP and Indel variation was measured by counting
positions in which an alternate allele, considered any base differing
from the Zv9 reference sequence, was present in only one of the three
sampled strains (red, yellow and blue quadrants in Fig. 1A). The
number of strain-specific variants ranged from a maximum of 1.35
million in WIK to a minimum of 0.62 million in TL, with a mean of
0.99±0.21 million SNPs (average strain-specific SNP ratio=1.34
SNPs/1 kb). The observed amount of relative strain-specific variation
is similar to estimates determined with limited SSLP markers (Coe
et al., 2009). In contrast to strain-specific variants, we defined shared
variants as any alternate alleles differing from the Zv9 reference
sequence and that were observed in more than one of the strains
surveyed. The majority of the shared variants we identified (2.32
million) were observed in all three strains (Fig. 1A, black), suggesting
that these genomic positions are commonly polymorphic in parental
wild zebrafish populations from which the laboratory strains were
derived. Some of the other shared variants demonstrated some strain
specificity being present in two of the three surveyed strains (Fig. 1A,
purple, orange and green quadrants). Of these, FLI (derived from EK)
and TL have the highest number of shared variants (1.21 million),
whereas WIK and TL contain the lowest (0.42 million).
As mentioned above, the frequency of a polymorphism often
dictates its utility. In genetic mapping, the likelihood of the presence
(high-frequency variants) or absence (low-frequency variants) of a
polymorphism will determine its value as a genetic marker.
Therefore, once grouped by strain specificity, variants were
classified by alternate allele read frequency into 10% intervals
(Fig. 1B-E). Alternate allele read frequencies (F) were simply
calculated by dividing the number of alternate allele reads by the
total number of sequence reads at a given position. As our
sequencing libraries for each strain were generated from a pool of 18
individuals, the alternate allele read frequencies reflect estimates of
the frequencies within the strain population rather than allele
frequencies calculated from individual genotyping. In both the TL
and WIK strains, the majority of strain-specific SNPs (60 and 55%,
respectively) had F≥0.5, whereas only 29% of FLI-specific SNPs
reached the same frequency (Fig. 1B). The higher number of fixed
(F=1) alternate alleles in the TL and WIK strains (19 and 13%,
respectively) compared with the FLI strain (5%) accounts for most
of the observed differences. The difference in the number of fixed
alleles might reflect a difference in population management, as FLI
has been propagated at the National Institutes of Health (NIH,
Bethesda, MD, USA) for 15 generations, whereas the TL and WIK
strains, surveyed in this study, were propagated for two generations
since acquisition from the Zebrafish International Resource Center
(ZIRC, Eugene, OR, USA). Strain-specific Indels had a similar
distribution when compared with strain-specific SNPs except that
Indels at F<0.3 were underrepresented (Fig. 1C).
For shared variants, alternate alleles that were fixed in all three
populations (Fig. 1D, black) accounted for the largest group in each
of the strains surveyed, ranging from 25% (TL) to 15% (FLI). This
observation suggests that the alleles found at these positions in the
Zv9 reference genome sequence are probably not the true
‘consensus’ laboratory zebrafish alleles, but rather TÜ-specific
1543
DEVELOPMENT
RESEARCH ARTICLE
RESEARCH ARTICLE
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
B
.94
.87
.42
TL SNPs (3.45 million)
FLI SNPs (3.87 million)
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
ALL
.4 < F < .5
FLI&TL
.5 < F < .6
.6 < F < .7
FLI&WIK
.7 < F < .8
.8 < F < .9
.9 < F < 1
F=1
0 5 10 15 20 25 30 35
% of Total Interstrain SNPs
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
ALL
.4 < F < .5
FLI&TL
.5 < F < .6
TL&WIK
.6 < F < .7
.7 < F < .8
.8 < F < .9
.9 < F < 1
F=1
0 5 10 15 20 25 30 35
% of Total Interstrain SNPs
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
.4 < F < .5
.5 < F < .6
.6 < F < .7
.7 < F < .8
.8 < F < .9
.9 < F < 1
F=1
0
ALL
FLI&TL
FLI&WIK
3
6
9
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
ALL
.4 < F < .5
FLI&WIK
.5 < F < .6
TL&WIK
.6 < F < .7
.7 < F < .8
.8 < F < .9
.9 < F < 1
F=1
0 5 10 15 20 25 30 35
% of Total Interstrain SNPs
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
.4 < F < .5
.5 < F < .6
.6 < F < .7
.7 < F < .8
.8 < F < .9
.9 < F < 1
F=1
0
TL Indels (486,285)
FLI Indels (533,242)
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
.4 < F < .5
.5 < F < .6
.6 < F < .7
.7 < F < .8
.8 < F < .9
.9 < F < 1
F=1
0
WIK SNPs (3.15 million)
Alternate Read Frequency
Alternate Read Frequency
E
.62
Alternate Read Frequency
Alternate Read Frequency
D
1.21
Alternate Read Frequency
2.32
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
.4 < F < .5
.5 < F < .6
.6 < F < .7
FLI (59,000)
.7 < F < .8
TL (47,950)
.8 < F < .9
.9 < F < 1
WIK (99,667)
F=1
10
0
5
15
20
% of Total Interstrain Indels
Alternate Read Frequency
1.35
C
0 < F < .1
.1 < F < .2
.2 < F < .3
.3 < F < .4
.4 < F < .5
.5 < F < .6
.6 < F < .7
FLI (882,876)
.7 < F < .8
TL (571,483)
.8 < F < .9
.9 < F < 1
WIK (1,251,908)
F=1
0
5
10
15
20
% of Total Interstrain SNPs
Alternate Read Frequency
Alternate Read Frequency
A
12
15
% of Total Interstrain Indels
WIK Indels (449,095)
ALL
FLI&TL
TL&WIK
3
6
9
12
15
% of Total Interstrain Indels
ALL
FLI&WIK
TL&WIK
3
6
9
12
15
% of Total Interstrain Indels
variants. Unlike shared SNPs, shared Indels did not show a bias
toward fixed alleles, but were more evenly distributed among the
frequency intervals in each strain (Fig. 1E). Again, most Indels
(59-70%) were present at some frequency in all three strains,
suggesting that these were probably common among the wild
zebrafish used to establish the laboratory strains.
In order to evaluate the spatial density of the identified variants,
the zebrafish genome was divided into 100-kb non-overlapping bins
and the number of SNPs per each bin (regardless of alternate allele
frequency) was measured. The resultant 100-kb bins were then
averaged by chromosome (Table 1). Genome-wide calculations
(average of each chromosome) of the total number of SNPs per
100 kb ranged from 382 (FLI) to 324 (TL), whereas the average
number of strain-specific SNPs was between 94 (WIK) and 47 (TL).
The ability to discriminate strains by SNP content (Table 1,
informative SNPs) was determined by counting the number of
1544
differential SNPs expected from pairwise crosses ignoring allele
frequency. With the exception of three chromosomes (16, 21 and
25), crosses between FLI and TL would yield the lowest number of
polymorphisms compared with crosses between FLI and WIK or
between TL and WIK. To provide a broader view of SNP density in
the zebrafish genome, the number of SNPs per 100-kb nonoverlapping bin was plotted along the length of each chromosome
(supplementary material Fig. S4).
Gene-associated variation in the zebrafish genome
A great advantage of the zebrafish model is the ability to conduct
large-scale forward-genetic screens. Most of these screens rely on
chemical mutagenesis with N-ethyl-nitrosourea (ENU) (SolnicaKrezel et al., 1994). With advances in NGS technologies, WGS has
been employed to find ENU-induced mutations responsible for
phenotypes of interest (Bowen et al., 2012; Obholzer et al., 2012;
DEVELOPMENT
Fig. 1. Natural SNP and Indel variation in laboratory zebrafish. (A) Venn diagram demonstrating the number of positions in the zebrafish genome with alternate
alleles specific to or shared between the FLI, TL and WIK strains as compared with the Zv9 reference genome. Numbers are in millions. See below for key to color
scheme. (B,C) The numbers of strain-specific SNP (B) and Indel (C) variants (red, blue and yellow color-coded categories of variants shown in A) as percentages
of total number of strain-specific variants (on the x-axis) and grouped by alternate allele read frequency (on the y-axis). The total number of strain-specific variants
for each strain is shown in parentheses to the right of the name of each strain. (D,E) The numbers of shared SNP (D) and Indel (E) variants ( purple, green, orange
and black color-coded categories of variants shown in A), as percentages of total number of shared variants (on the x-axis) and grouped by observed alternate
allele read frequency (on the y-axis). The total number of variants is shown above each graph and the color scheme is shown to the right. The color code in all
panels denotes which strain or strains contain(s) a variant at a given position (at any frequency) relative to the Zv9 reference: Red, present in only FLI; blue,
present in only TL; yellow, present in only WIK; purple, present in FLI and TL, but not in WIK; green, present in TL and WIK, but not in FLI; orange, present in FLI
and WIK, but not in TL; black, present in all three strains. F, alternate allele read frequency.
RESEARCH ARTICLE
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
Table 1. Average number of SNPs per 100-kb window
Total number of SNPs
Strain-specific number of SNPs
Informative number of SNPs
Chromosome
FLI
TL
WIK
FLI
TL
WIK
F×T
F×W
T×W
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
AVG
SD
369
371
387
299
441
383
371
360
394
401
395
355
375
369
395
335
402
370
404
361
416
364
420
370
432
382
30
320
321
158
288
317
345
327
330
328
321
362
279
330
325
347
288
368
305
356
335
330
336
364
332
380
324
42
367
285
158
308
349
351
337
334
373
363
398
258
393
346
344
367
353
379
390
375
327
287
401
337
391
343
52
66
93
69
43
57
69
66
67
70
81
49
80
61
61
89
58
75
79
67
66
83
87
62
71
76
70
12
45
58
41
43
50
52
43
57
48
35
30
49
38
41
66
40
60
48
50
54
36
66
38
53
50
48
9
110
70
92
88
99
97
87
92
93
94
123
49
112
84
92
114
89
115
102
101
82
61
108
93
112
94
17
209
237
209
173
207
69
204
224
229
221
157
226
208
185
255
367
228
253
223
234
207
251
202
221
391
224
59
291
297
285
243
273
97
282
255
273
296
309
267
292
252
311
221
301
314
294
289
303
291
306
295
221
274
44
299
278
297
254
283
97
279
300
295
284
303
261
292
255
311
300
306
317
308
291
113
276
312
298
333
278
54
AVG, average; SD, standard deviation. F×T, FLI×TL strain; F×W, FLI×WIK strain; T×W, TL×WIK strain.
Voz et al., 2012; Miller et al., 2013). However, to recognize a
causative mutation, other mutation candidates need to be eliminated
from consideration. We examined our natural SNP and Indel
UTR
40
ncRNA
20
F T W F T W F T W
F T W F T W F T W
Strain
Strain
Stop Variants
Charge
Polarity
800
700
600
500
400
300
200
100
F
SYN
NON
F T W F T W
Strain
Indel Variants
G
200
Splice Variants
1000
SD
FS
NS
800
150
RT
100
IF
50
SNPs
E
SNPs X 1000
60
SNPs
Like
80
Codon Variants
80
70
60
50
40
30
20
10
SA
600
400
200
F T W F T W F T W
F T W F T W
F T W F T W
F T W F T W
Strain
Strain
Strain
Strain
FLI
TL
WIK
FLI & TL
FLI & WIK
TL & WIK
ALL
Fig. 2. Spectrum of natural genetic variation in laboratory zebrafish. The number of SNPs or Indels per category. (A) Unexpressed variants, including
intergenic (Inter), intronic (Intron) and within 5 kb of a transcript (Proximal) variants. (B) Transcript variants, including those falling in exons, non-coding RNA
(ncRNA), and untranslated regions (UTR). (C) Codon variants, divided into synonymous (SYN) and non-synonymous (NON) changes. (D) Missense variants,
divided into conservative changes, e.g. nonpolar-to-nonpolar (Like), charge changes, e.g. polar-to-positive (Charge) and polarity changes, e.g. polar-to-nonpolar
(Polarity). (E) Stop codon variants, divided into nonsense (NS) and read-through (RT) alleles. (F) Indel variants within coding sequences, divided by whether they
result in frameshift (FS) or in-frame (IF) alleles. (G) Splice site-altering variants, including those that alter splice donor dinucleotides (SD) and those that alter
splice acceptor dinucleotides (SA). Color scheme used is as in Fig. 1, and is also noted at the bottom of this figure.
1545
DEVELOPMENT
Proximal
Missense Variants
8
7
6
5
4
3
2
1
EXON
Inter
Intron
C
Transcript Variants
100
SNPs
SNPs X 1000
D
B
Unexpressed Variants
350
300
250
200
150
100
50
SNPs x 1000
SNPs X 10,000
A
variants for their presence in coding regions of the genome and their
putative effect, if any, on the encoded peptides. Natural SNP and
Indel variation was categorized using Annovar (Fig. 2A-G) (Wang
RESEARCH ARTICLE
et al., 2010). As expected, the vast majority of SNP variants were
located in either intergenic or intronic sequence (Fig. 2A). A mean
of 8.4×104 (s.e.m.=0.5×104) SNP positions were located in the
exons for each strain (Fig. 2B, EXON). Of the exonic variations,
about one quarter (23.4%) caused non-synonymous amino acid
changes in each strain (Fig. 2C). Many of these variants would be
considered strong mutation candidates if located in crucial regions
identified by meiotic mapping, as they include changes in polarity
or charge of amino acid side chains (Fig. 2D), nonsense and readthrough changes (Fig. 2E), and frameshift and non-frameshift
insertions/deletions (Fig. 2F). Similarly, some of the SNP variants
were located in splice acceptor and donor sites, resulting in potential
aberrant splicing (Fig. 2G). In order to determine the distribution of
non-synonymous mutations among genes, the number of nonsynonymous changes per kb of exonic sequence was calculated for
each Refseq gene annotation in the zebrafish genome. When all
14,484 genes are considered, an average of 56% of the total genes
showed ≥1 non-synonymous change per kb of exonic sequence
among the surveyed strains (supplementary material Fig. S5).
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
A
TL
NIH
0.78
Shared
3.22
UPenn
1.05
B
WIK
NIH
1.50
Shared
2.70
UPenn
1.46
Alternate alleles are present in independent population
isolates
Our data on genetic variation in laboratory zebrafish were organized
into an easy-to-use website termed ‘SNPfisher’ (snpfisher.nichd.
nih.gov). For acquiring and filtering variants in a gene or genomic
region of interest, the SNPfisher variant report compiler was
developed (Fig. 4A). Users define a genomic region of interest using
the UCSC genome browser coordinate format or provide a Refseq
gene name in the query box (Fig. 4A). Optional filters located below
the main query box allow user-defined criteria for the variant
frequency and the minimum read depth (Fig. 4A, dashed red and
blue boxes, respectively). An example of the SNPfisher variant
report compiler annotation is shown in Fig. 4B. This report includes
the genotype of the population (GT), read depth at the variant
1546
<.1
.1 to <.2
.2 to < .3
.3 to < .4
.4 to < .5
.5 to < .6
.6 to < .7
.7 to < .8
.8 to < .9
.9 to < 1.0
1.0
0
20
40
60
80
100
% NIH SNPs detected
Fig. 3. Variant sharing between independent populations of TL and WIK.
(A,B) Venn diagrams showing the number of SNPs shared (overlapping areas)
or not shared (non-overlapping areas) between independent isolates of the
TL (A) and WIK (B) strains maintained at the NIH (leftmost circle) and
UPenn (rightmost circle). (C) A stacked bar graph showing the percentage of
total NIH SNPs also detected in the UPenn SNP dataset (x-axis), grouped by
frequency category ( y-axis). Frequency is calculated by dividing the number of
alternate allele reads per total reads from the NIH strains. Values for the TL and
WIK strains are shown in blue and yellow, respectively.
position (DP), the allele distribution (AD) and the alternate allele
read frequency (FQ=alternate allele reads divided by the total
number of reads) for each of the three surveyed strains (FLI, TL,
WIK) (Fig. 4B). To expedite the use of these variants, each was
provided a link to the ‘Primer3’ web interface to generate primers
for PCR amplification (Rozen and Skaletsky, 2000). To further
facilitate the use of these variants in applications like allele-specific
PCR or genetic fine mapping, each variant was evaluated for
restriction fragment length polymorphisms (Fig. 4B). Likewise, to
aid in the selection of variants as molecular markers, an external
data hyperlink (EX_Data; Fig. 4B) was built into the SNPfisher
variation report to assess the presence or absence of variants in other
laboratory populations (Fig. 4C). Additionally, external variant
DEVELOPMENT
Zebrafish variant report compiler tool
C
ALT Read Frequency
The relative amount of genetic drift from the founders of a given
strain will vary among different isolates maintained in independent
laboratories. To determine the utility of our variation data for the
larger zebrafish community, we compared our variation data
( populations maintained by the Weinstein laboratory at the NIH)
with independent variation data obtained from populations
maintained by the Granato laboratory [University of Pennsylvania
(UPenn), Philadelphia, PA, USA]. The NIH populations of TL
and WIK were both acquired from the ZIRC and maintained at NIH
for two generations. The UPenn and ZIRC populations of TL and
WIK were derived from the same parental populations kept by the
Max Planck Institute for Developmental Biology (Tübingen,
Germany), but have been propagated independently for ∼10-15
generations. This comparison was restricted to SNP positions that
had at least three reads in both the NIH and UPenn data. In this
comparison, the TL and WIK strains shared 3.22 and 2.70 million
SNPs (61% and 48% of total), respectively (Fig. 3A,B). The shared
SNP positions were then quantified and grouped based on variant
frequency in our dataset, as shown in Fig. 3C. Over 99% and 76% of
the fixed alleles (1.0 value on y-axis) in the TL and WIK strains,
respectively, were also detected in the UPenn dataset, suggesting
that most of the SNP data at higher frequencies will be common
among different isolates of zebrafish for both the TL and WIK
strains (Fig. 3). The higher percentage of SNPs detected in the TL
strain suggests a lower level of genetic diversity in this strain
compared with the WIK strain.
RESEARCH ARTICLE
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
A
B
Fixed TLVariants
D
1000
150
800
120
600
90
400
60
200
30
<= 1.0 0.8 0.6 0.4 0.2 0.0
WIK Alt Read Frequency
E
10
20
30
40
50
Read Depth
Fixed WIK Variants
C
1000
200
800
150
600
100
400
50
200
<= 1.0 0.8 0.6 0.4 0.2 0.0
10
TL Alt Read Frequency
20
30
40
50
Read Depth
annotations can be queried directly by launching the ‘External Data’
hyperlink at the SNPfisher homepage. This link opens a query box
to retrieve the desired external variation data by refseq gene name or
genomic coordinates.
In order to demonstrate the ease and utility of the SNPfisher
variant report compiler, variants were evaluated for their
informativeness in the ∼800 kb crucial region that has been
reported for the cloche mutation (Leshchiner et al., 2012).
Hypothetically, if recombination in the cloche region were to be
determined by segregation of TL versus WIK alleles, the allele
frequency and depth filters could quickly prioritize variants to serve
as molecular markers. An unfiltered search of the region returns
4683 variants. Requiring fixation of variant alleles (alt. read
frequency=1.0) in TL or WIK decreases the number of variants to
837 and 853, respectively. Requiring absence of TL variants in WIK
and vice versa drops the number of variants to 160 and 145 for TL
and WIK variants, respectively (left plots in Fig. 4D,E). Additional
filtering by read depth, which is proportional to the accuracy of the
alternate allele read frequency estimate, quickly reduces the number
of candidate variant markers (right plots in Fig. 4D,E). Thus, the
SNPfisher variant report compiler not only documents variants, but
also prioritizes them to suit a user’s needs.
Genome-wide variant resources
To facilitate genome-wide approaches, like filtering strain variants
from NGS-based positional cloning projects, the SNPfisher site
1547
DEVELOPMENT
Fig. 4. SNPfisher homepage and variant report compiler tool. (A) Screen shot of the SNPfisher homepage featuring the SNPfisher Variant Report Compiler.
Users can define frequency values (dashed red box) and minimum read depth values (dashed blue box) for each strain in any gene or genomic region entered into
the query box (white box in center). The drop-down filtering menus include ‘=’, ‘<’, ‘>’, ‘≤’ and ‘≥’ to adjust variant frequencies. (B) A variant annotation
generated by the SNPfisher Variant Report Compiler homepage. The Primers column contains links to the ‘Primer3’ web interface to generate primers for
amplification of the variant. The genotype (GT) value corresponds to genotype of the population of fish surveyed (0/0=ref. allele homozygote, 0/1=alt. allele
heterozygote, 1/1=alt. allele homozygote). The allele distribution (AD) is composed of the number of ref. allele reads and then alt. allele reads separated by
commas, respectively. Alt. allele read frequencies (FQ) were calculated by dividing the alt. allele reads by the total number of reads at a given position. Variants
encoding RFLPs are documented in the RE_Changes column. The creation (+) or destruction (−) of a site is indicated prior to the restriction enzyme name. The
number in parentheses following the restriction enzyme name indicates the number of restrictions recognition sites in the 800 bp surrounding the variant
(400 bp upstream and downstream) in the Zv9 reference genome sequence. (C) An example of the variant annotation provided by the Ex_Data link in the
SNPfisher report. The total read depth (DP), ref. allele read depth (WT), alt. allele read depth (MUT) and alt. allele read frequency (FQ) values represent the
cumulative read sums calculated from published datasets. (D,E) Scatterplots demonstrate TL and WIK variant filtration in the cloche crucial region. The
number of fixed TL variants ( y-axis in left plot of D) decreases by filtering for lower alternate (Alt) allele read frequencies in WIK (x-axis). Increasing read depth
(x-axis in right plot of D) decreases fixed TL variants absent in WIK (x-axis). Similar results are obtained for WIK variants in the cloche region (E).
RESEARCH ARTICLE
provides our datasets in variant call format files (VCF). These files
are available from the ‘Tracks and Datasets’ page of the SNPfisher
website (Fig. 5A, column 2). The SNPfisher VCF contains 7.86
million variants. Additionally, VCFs partitioned by strain
distribution are also provided for download. To demonstrate its
filtering capability, the SNPfisher dataset was used to filter the
linkage region of the published ca1 mutant (Leshchiner et al.,
2012). The SNPTrack pipeline, after applying its own inherent
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
variant filtration, identifies ten missense mutations as the most likely
causative mutations. SNPfisher filtration reduces the number of
candidates to six, having four annotated as natural variants.
In order to enhance genome-wide variant lists, we combined our
variants along with previously published variants (Bowen et al.,
2012; Obholzer et al., 2012). Our dataset, as well as the others, were
generated from sequencing libraries of pooled genomic DNA
samples. Consequently, the alternate allele read frequency is an
A
Fig. 5. SNPfisher tracks and dataset page. (A) Screen shot of SNPfisher ‘Tracks and Datasets’ page containing UCSC genome track links (columns 1 and 3)
and downloadable datasets in VCF (column 2) and CVF format (column 4). The bigBed file hyperlinks (Combined FLI/TL/WIK, FLI, TL, WIK) display variation data
and the bigWig files (FLI, TL and WIK) show base-by-base sequencing coverage for each strain in the UCSC genome browser. CVF variants were also
formatted for display in the UCSC genome browser (column 3). (B) Screen shot of the brcc3 locus (Miskinyte et al., 2011) in the UCSC genome browser, in
which SNP and Indel variation is demonstrated with the Combined FLI/TL/WIK Variant Track (top), the FLI Variant Track (middle) and the FLI Coverage Track
(bottom). Variant annotations are formatted by Ref. Allele_Alt. Allele:Read Depth_Alt. Allele Read Frequency. In the Combined FLI/TL/WIK Variant Track,
the strain distribution is annotated by color (FLI, red; TL, blue; WIK, yellow; FLI&TL, purple; FLI&WIK, orange; TL&WIK, green; ALL, black), and depth and alt allele
read frequency are calculated from reads from all three strains.
1548
DEVELOPMENT
B
estimate of the alternate allele frequency in the population of fish
represented in the pooled genomic sample. Therefore, the available
datasets were combined with our variants in a format we have named
consortial variant format (CVF). CVFs contain the chromosome or
contig (CHROM), position (POS), reference allele (REF), alternate
allele (ALT), reference read depth (REF_DP), alternate allele read
depth (ALT_DP), total read depth (DP) and alternate allele read
frequency (AF). All of the depth values are sums calculated from the
collective datasets available. CVFs for the AB, TL, TÜ and WIK
strains, containing 5064481, 6994338, 4094439 and 9661402
variants, respectively, are available for download from the
SNPfisher ‘Tracks and Datasets’ page (Fig. 5A, column 4).
SNPfisher genome browser displays
For integrated inquiries with other genome annotations, SNPfisher
variants were formatted for display in the UCSC genome browser
(Kent et al., 2002) and accessed through the ‘Tracks and Datasets’ link
at the SNPfisher homepage (Fig. 5A, column 1). SNPfisher variants
were organized into bigBed files and the sequencing coverage of each
base of the zebrafish genome was provided in bigWig files. Variant
annotation consists of the Zv9 reference allele, the alternate allele, the
read depth and the frequency of the alternate allele per total alleles, as
demonstrated in Fig. 5B. The frequency information for the Combined
FLI/TL/WIK track is calculated from the sum of the alternate allele
reads divided by the total number of reads from all strains. For the
Combined track, the strains are represented by color, as shown in
Fig. 5B (top). SNPfisher also provides a convenient means to scan
short oligos for SNPs and Indels. Users must first launch the links for
the strain variation of interest (bigBeds) on the ‘Tracks and Dataset’
page. Then, from the SNPfisher homepage, launching the ‘Blat a
Sequence’ link will provide a query box for the UCSC Blat tool to
search any oligo sequence (Fig. 4A, left) (Kent, 2002). To show the
utility of using this tool, a genome-wide survey of splice morpholino
(spMO) target site and predicted CRISPR site variation was
conducted. Using the SNPfisher variation dataset alone, at least one
variant was identified in 48% and 21% of spMO (498,064) and
predicted CRISPR (4.28 million) target sites, respectively.
DISCUSSION
Historically, the low number of identified SNPs has hampered their
use in the zebrafish field (Stickney et al., 2002; Guryev et al., 2006;
Bradley et al., 2007). The recent use of WGS and RNA-seq to clone
zebrafish mutations has generated a substantial amount of SNP data
in the last few years (Bowen et al., 2012; Leshchiner et al., 2012;
Obholzer et al., 2012; Voz et al., 2012; Miller et al., 2013).
However, the annotation and relative ease of accessing SNP
variation data remain scant and insufficient, respectively. Similarly,
the utility of the known SNPs is often limited by the absence of
annotations for strain specificity and allele frequency.
In this study, we sampled the natural genetic variation within and
between three laboratory zebrafish strains by WGS of pooled DNA
samples, encompassing 18 individuals per strain. Each sample was
sequenced to ∼16× coverage, yielding an average of 5.04 million
SNPs per strain when compared with the Zv9 reference strain. One
drawback of using pooled DNA samples is the inability to directly
measure allele frequencies for reference and alternate alleles.
However, we assumed that each of the 36 alleles in our samples had
an equal probability of being sequenced. Consequently, we can
estimate the allele frequency from the ratio of alternate allele reads to
reference allele reads at a given position. Clearly, the accuracy of our
estimated allele frequencies will be proportional to the total read
depth per position. A similar proportional relationship will exist
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
between the observed strain specificity and total read depth, as
alternate or reference alleles are less likely to be missed at higher
read depths. Importantly, the SNPfisher reporter tool enables
researchers to easily filter by minimum read depth, allowing each
researcher to choose their confidence threshold.
For both the TL and WIK strains, we directly compared the variation
in our populations to independent populations maintained in the
Granato laboratory at the University of Pennsylvania to address the
applicability of using our SNP data in other populations. When
alternate allele frequency is considered, almost all (>99%) of the TL
variant positions fixed in the NIH population (1.25 million) are
detected in the UPenn population, suggesting that our variation data
represent the TL strain well. At the same alternate allele frequency,
fewer WIK variant positions in the NIH population (0.90 million; 77%)
are detected in the UPenn population. Still, the vast majority of fixed
variants are detected, suggesting that the variation observed in the WIK
population at NIH can be applied reasonably well to other WIK isolates.
In order to facilitate access and use of our dataset, we generated the
SNPfisher tool and formatted our data for display in the UCSC
genome browser. SNPfisher provides a user-friendly, web-based
interface for the acquisition of SNPs and Indels in any region of the
zebrafish genome. Users can supply minimums for read depth and
allele frequency in any or all strains surveyed (EK, TL or WIK) to
quickly ascertain SNPs of interest. Additionally, the display of our
SNP data in the UCSC genome browser facilitates integrated
inquiries of SNP variation to augment primer selection and target
site selection for TALENs, CRISPRs or morpholinos (Nasevicius
and Ekker, 2000; Huang et al., 2011; Sander et al., 2011; Chang et al.,
2013; Hwang et al., 2013). Variation in intended target sites can be a
significant issue in the design of effective tools for site-directed
mutation or gene knockdown. For example, unique CRISPR target
sites (for the Schizosaccharomyces pombe Cas9 system) have been
predicted for the entire zebrafish genome (http://www.genomeengineering.org/crispr/?page_id=41). Examination of the 4.28
million predicted CRISPR target sites revealed that 21% contained
SNPs annotated by SNPfisher. Similarly, we evaluated over 498,000
splice morpholino target sites and determined that 48% had at least 1
annotated variant in SNPfisher. Taken together, these observations
argue for the incorporation of variation data into sequence-based
molecular biology reagent design, which is now easy with SNPfisher.
The SNP variation that we report in this study will enhance the
identification of genetic lesions in forward-genetic approaches. Our
data will improve the discrimination between causative mutations
versus neutral polymorphisms. In addition, the identification of SNP
variation in studies like ours will enable genome-wide association
studies to tackle complex traits in the zebrafish. Hypothetically, an
ENU-mutagenesis screen could be conducted to identify a family
with an extreme spectral phenotype that exhibits non-Mendelian
inheritance. Given the fecundity of zebrafish, a genome-wide
association scan could be performed within the family of interest
comparing affected to wild-type siblings, eliminating signals from
population stratification without compromising sample size.
MATERIALS AND METHODS
Ethical statement
Zebrafish work was performed in accordance with NIH policy and approved
by the Institutional Animal Care and Use Committee.
Zebrafish strains and gDNA isolation
NIH isolates. Tüpfel long fin (TL) and WIK (Rauch et al., 1997) zebrafish
strains were purchased from the ZIRC and propagated for two generations.
The Tg( fli1a-eGFP)y1 strain (Lawson and Weinstein, 2002) is derived from
1549
DEVELOPMENT
RESEARCH ARTICLE
RESEARCH ARTICLE
a transgenic insertion into the EkkWill (EK) strain and has been maintained
at NIH for 14 generations, including several backcrosses to wild-type EKs.
For each strain, the caudal fins were clipped from 18 individual adult
zebrafish (nine female and nine male), and the clips were pooled into a
single sample from which gDNA was isolated using standard techniques
(Strauss, 2001).
UPenn isolates. TL and WIK samples contained gDNA isolated and
pooled from the somatic tissue of 26 and 27 ENU-mutagenized adult males,
respectively. Males were selected prior to mutagenesis based on the absence
of startle response, performance and habituation behaviors.
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
patterns for the list of enzymes available on the commercial website of New
England Biolabs (NEB).
SNP density calculations
The size of the zebrafish genome (Zv9) is reported to be 1.506 Gb (Howe
et al., 2013). Repeatmasker masked 769 Mb of the Zv9 genome as repetitive
sequence (51%), leaving 737 Mb of unique sequence. The size of the human
genome is 3.14 Gb, with Repeatmasker marking 1.47 Gb as repetitive
sequence. The remaining 1.66 Gb of unique sequence was used for density
calculations.
Whole-genome small insert library method and sequencing
ca1 Mutant filtration
The NIH Intramural Sequencing Center (NISC) performed library
preparation and WGS. Five micrograms (μg) of genomic DNA was
sheared to ∼500 bp using a Covaris E210 with the following settings:
duty cycle 20%; intensity 4; cycles 200. The DNA was size-selected for a
range of 300-600 bp on a Pippin Prep (Sage Science) 1.5% cassette. The
library was constructed on a SPRI-TE Robot (Beckman Coulter) using
SPRI-TE reagents and Illumina Paired-End DNA Sample Prep Oligo
Only Kit. The resulting library was size-selected on a Pippin Prep with a
1.5% cassette selecting for 440-560 bp. To ensure that the library was not
overamplified, test amplification was performed in which aliquots of a
PCR reaction were removed every two cycles from 4-16. These aliquots
were evaluated on a 2% agarose gel and an optimal cycle number was
selected for subsequent large-scale amplification in which unique
barcodes were added to each library. For these libraries 12-14 cycles
were selected. Amplification reactions were cleaned up using two rounds
of Agencourt AMPure Beads. Library aliquots were pooled and the pool
was evaluated on a MiSeq indexed run. The de-multiplexed read numbers
were used to produce a normalized pool for sequencing on a HiSeq2000.
The five-member pool was run in one lane to generate paired-end 101
base reads. Additional reads were generated for three of the libraries (FLI,
WIK and TL). These three were run in individual lanes under the same
run conditions. Data were processed using RTA1.12.4.2 or 1.13.48 and
CASAVA 1.8.2.
SRR453533 and SRR453534 were downloaded from the Sequence Read
Archive (http://trace.ncbi.nlm.nih.gov/Traces/sra/sra.cgi?view=search_obj)
and converted to FASTQ format with the fastq-dump tool in the SRA
Toolkit (Wheeler et al., 2008). The resultant FASTQs were uploaded to
SNPTrack for linkage analysis and mutation screening (Leshchiner et al.,
2012). SNPs from the linked region (chr3:21.4-33.2 Mb), as indicated by
the HMM score, were downloaded from SNPTrack if they met each of the
following criteria: (1) homozygous in the mutant pool; (2) heterozygous in
the wild-type pool; (3) not annotated in the SNPTrack SNP database; and (4)
not repetitive. The most likely causative mutations (ten missense mutations)
of these SNPs were evaluated for presence in SNPfisher.
WGS alignment and post-processing
Reads in FASTQ format obtained from Illumina HiSeq sequencing were
aligned to the Zv9 assembly (Ensembl) for each sample using Bowtie2
(Langmead and Salzberg, 2012). Following alignment, the BAM files were
treated according to the GATK best practices workflow (https://www.
broadinstitute.org/gatk/guide/best-practices) (McKenna et al., 2010;
DePristo et al., 2011). Read clipping was performed and duplicates were
marked and removed with Picard (http://picard.sourceforge.net). In addition,
reads with mapping quality under 30 were discarded using Samtools and a
custom Perl script (Li et al., 2009). Finally, SNPs and Indels were called
across all datasets in parallel using the GATK unified genotyper (McKenna
et al., 2010; DePristo et al., 2011).
Splice morpholino and CRISPR target site polymorphisms
CRISPR target sites were downloaded from the CRISPR Genome
Engineering Resources website (http://www.genome-engineering.org/
crispr/) via the UCSC genome browser. All annotated exon start and stop
coordinates in Zv9 were extracted from Ensembl gene annotations
downloaded from the UCSC genome browser (http://hgdownload.soe.
ucsc.edu/goldenPath/danRer7/database/ensGene.txt.gz). spMO target sites
were defined as 50 bp (25 upstream, 25 downstream) surrounding the last
base of the exon of splice donors and the first base of the exon of splice
acceptors. A custom script was written to generate a non-redundant bed of
spMO target sites from the extracted exon target sites. The intersect function
of bedtools was used to identify spMO and CRISPR target sites that contain
SNPfisher variants (Quinlan and Hall, 2010).
SNPfisher tool
All SNPs and Indels passing all filtering criteria along with predicted
restriction enzyme changes were compiled in a simple database accessible
using a web interface (snpfisher.nichd.nih.gov). Simple filtering criteria are
implemented in the submission form, including restriction digest
requirements, allele frequencies and associated read depths for each or
any of the strains. Sequencing data and mutation calls were compiled into
UCSC genome browser tracks for easy viewing of both mutations and
sequencing coverage for each genetic background.
SNPs and Indels were split and flagged according to manually defined filters
as follows (FLAG: Filter logic): InDel: SNP mutations occurring within
10 bp of Indels; MapQuality: Mapping Quality <40; LowQual: QUAL <30;
HighDepth: Depth >120; LowDepth: Depth <3; SNPCluster: ≥3 SNPs in a
10-bp window. Indels were flagged according to the following: LowQD:
Quality depth ratio (QD) <2. Flagged mutations were discarded from
consideration. Additionally, called mutations were required to pass data for
all three datasets; a minimum of three reads must exist in each of the three
strains for any single mutation to be considered. Mutations mapped to
repetitive elements and low-complexity regions were removed using a
Repeatmasker annotation file of the Zv9 sequence (http://repeatmasker.org).
The SNPs were categorized by location and consequence using Annovar
(Wang et al., 2010). Basic statistics regarding gene position and resulting
mutation changes were compiled using a custom Perl script. SNPs
introducing or eliminating restriction enzyme recognition sites were
predicted using a custom Perl script, using restriction enzyme mutation
1550
Binary alignment maps (BAMs) for each wild-type strain sequenced by
Bowen et al. were downloaded from the Resource page of the Harris Lab
website (Boston Children’s Hospital, Boston, MA, USA; http://fishbonelab.
org/harris/Resources.html) (Bowen et al., 2012). BAMs divided by
chromosome were concatenated into a single BAM per each strain.
Variants were recalled and filtered independently, not with other strains,
according to the criteria described by Bowen et al., with the exception that a
low-depth filter was removed due to low genome coverage.
VCF files for the wild-type strains sequenced by Obholzer et al. were
downloaded from the MegaMapper page of the Megason Lab website
(https://wiki.med.harvard.edu/SysBio/Megason/MegaMapper; Obholzer
et al., 2012). A Repeatmasker annotation file for the Zv9 assembly was
downloaded from http://hgdownload.soe.ucsc.edu/goldenPath/danRer7/
database/rmsk.txt.gz and re-ordered to match the contig order in the Zv9
fasta downloaded from the Wellcome Trust Sanger Institute website
(ftp://ftp.sanger.ac.uk/pub/zfish/assembly/Zv9/Zv9_toplevel.fa.gz). The reordered Repeatmasker file was used as a mask with the GATK
DEVELOPMENT
Consortial variant format
Filtering and annotation of SNPs and Indels
VariantFiltration tool to flag and then remove repeatmasked variants from
the original VCFs created by Obholzer et al. A bash script was then written
to calculate the QD values for each variant and remove any variants with a
QD<2.0.
CVFs were constructed from VCFs from each dataset. CVFs are
composed of the chromosome or contig (CHROM), position in the Zv9
genome reference assembly (POS), Zv9 reference allele (REF), alternate
allele (ALT), cumulative Zv9 reference allele read depth (REF_DP),
cumulative alternate allele read depth (ALT_DP), cumulative total read
depth (DP) and the cumulative alternate allele read depth (AF). The allele
read depths and allele read frequencies are referred to as cumulative, as they
are the sum of the values from each independent dataset. Inclusion in the
CVFs was restricted to biallelic variants, as REF_DP and ALT_DP values
could not be calculated from DP4 values in the INFO fields of polyallelic
variants in the datasets of Bowen et al. (2012) and Obholzer et al. (2012).
Data access
All SNP and Indel variants that passed all of the filters imposed by
the authors are available at the snpfisher.nichd.nih.gov website. Raw
seq files will be submitted to Sequence Read Archive (SRA) at
http://www.ncbi.nlm.nih.gov/Traces/sra/sra.cgi upon acceptance
under the accession numbers SRS863087, SRS863089, SRS863090.
Acknowledgements
We would like to thank Alice Young (NISC) for providing the methods for genomic
library construction and Illumina sequencing performed in this study. We would also
like to thank Matthew Breymaier for providing website and server expertise. This
study utilized the high-performance computational capabilities of the Helix Systems
and Biowulf cluster at the NIH (http://helix.nih.gov).
Competing interests
The authors declare no competing or financial interests.
Author contributions
M.G.B. and B.M.W. conceived and designed the experiments. M.G.B. performed
the experiments. Data analysis was conducted by M.G.B., J.R.I. and J.A.E. M.G.B.
and B.M.W. developed the SNPfisher website. J.R.I. wrote the code for the
SNPfisher website. K.C.M. and M.G. provided WGS data for comparison. M.G.B.
and B.M.W. wrote the paper.
Funding
This work was supported by the intramural program of the Eunice Kennedy Shriver
National Institute of Child Health and Human Development (NICHD) and by the
Fondation Leducq (Transatlantic Network of Excellence for the Identification of
Novel Genetic Targets in Hemorrhagic Stroke) to B.M.W. This work was also
supported by a National Institutes of Health grant [MH092257] to M.G. Deposited in
PMC for release after 12 months.
Supplementary material
Supplementary material available online at
http://dev.biologists.org/lookup/suppl/doi:10.1242/dev.118786/-/DC1
References
Abecasis, G. R., Altshuler, D., Auton, A., Brooks, L. D., Durbin, R. M., Gibbs,
R. A., Hurles, M. E. and McVean, G. A. (2010). A map of human genome
variation from population-scale sequencing. Nature 467, 1061-1073.
Abecasis, G. R., Auton, A., Brooks, L. D., DePristo, M. A., Durbin, R. M.,
Handsaker, R. E., Kang, H. M., Marth, G. T. and McVean, G. A. (2012). An
integrated map of genetic variation from 1,092 human genomes. Nature 491, 56-65.
Bowen, M. E., Henke, K., Siegfried, K. R., Warman, M. L. and Harris, M. P. (2012).
Efficient mapping and cloning of mutations in zebrafish by low-coverage wholegenome sequencing. Genetics 190, 1017-1024.
Bradley, K. M., Elmore, J. B., Breyer, J. P., Yaspan, B. L., Jessen, J. R., Knapik,
E. W. and Smith, J. R. (2007). A major zebrafish polymorphism resource for
genetic mapping. Genome Biol. 8, R55.
Chang, N., Sun, C., Gao, L., Zhu, D., Xu, X., Zhu, X., Xiong, J.-W. and Xi, J. J.
(2013). Genome editing with RNA-guided Cas9 nuclease in zebrafish embryos.
Cell Res. 23, 465-472.
Coe, T. S., Hamilton, P. B., Griffiths, A. M., Hodgson, D. J., Wahab, M. A. and
Tyler, C. R. (2009). Genetic variation in strains of zebrafish (Danio rerio) and the
implications for ecotoxicology studies. Ecotoxicology 18, 144-150.
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
DePristo, M. A., Banks, E., Poplin, R., Garimella, K. V., Maguire, J. R., Hartl, C.,
Philippakis, A. A., del Angel, G., Rivas, M. A., Hanna, M. et al. (2011). A
framework for variation discovery and genotyping using next-generation DNA
sequencing data. Nat. Genet. 43, 491-498.
Guryev, V., Koudijs, M. J., Berezikov, E., Johnson, S. L., Plasterk, R. H. A., van
Eeden, F. J. M. and Cuppen, E. (2006). Genetic variation in the zebrafish.
Genome Res. 16, 491-497.
Haffter, P., Odenthal, J., Mullins, M. C., Lin, S., Farrell, M. J., Vogelsang, E.,
Haas, F., Brand, M., van Eeden, F. J. M., Furutani-Seiki, M. et al. (1996).
Mutations affecting pigmentation and shape of the adult zebrafish. Dev. Genes
Evol. 206, 260-276.
Harper, C. and Lawrence, C. (2011). The Laboratory Zebrafish. Boca Raton, FL:
CRC Press.
Howe, K. Clark, M. D. Torroja, C. F. Torrance, J. Berthelot, C. Muffato, M.
Collins, J. E. Humphray, S. McLaren, K. Matthews, L. et al. (2013). The
zebrafish reference genome sequence and its relationship to the human genome.
Nature 496, 498-503.
Huang, P., Xiao, A., Zhou, M., Zhu, Z., Lin, S. and Zhang, B. (2011). Heritable
gene targeting in zebrafish using customized TALENs. Nat. Biotechnol. 29,
699-700.
Hwang, W. Y., Fu, Y., Reyon, D., Maeder, M. L., Tsai, S. Q., Sander, J. D.,
Peterson, R. T., Yeh, J.-R. J. and Joung, J. K. (2013). Efficient genome editing in
zebrafish using a CRISPR-Cas system. Nat. Biotechnol. 31, 227-229.
Kent, W. J. (2002). BLAT–the BLAST-like alignment tool. Genome Res. 12,
656-664.
Kent, W. J., Sugnet, C. W., Furey, T. S., Roskin, K. M., Pringle, T. H., Zahler, A. M.
and Haussler, D. (2002). The human genome browser at UCSC. Genome Res.
12, 996-1006.
Knapik, E. W., Goodman, A., Ekker, M., Chevrette, M., Delgado, J., Neuhauss, S.,
Shimoda, N., Driever, W., Fishman, M. C. and Jacob, H. J. (1998). A
microsatellite genetic linkage map for zebrafish (Danio rerio). Nat. Genet. 18,
338-343.
Langmead, B. and Salzberg, S. L. (2012). Fast gapped-read alignment with Bowtie
2. Nat. Methods 9, 357-359.
Lawson, N. D. and Weinstein, B. M. (2002). In vivo imaging of embryonic vascular
development using transgenic zebrafish. Dev. Biol. 248, 307-318.
Leshchiner, I., Alexa, K., Kelsey, P., Adzhubei, I., Austin-Tse, C. A., Cooney,
J. D., Anderson, H., King, M. J., Stottmann, R. W., Garnaas, M. K. et al. (2012).
Mutation mapping and identification by whole-genome sequencing. Genome Res.
22, 1541-1548.
Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G.,
Abecasis, G. and Durbin, R. (2009). The sequence alignment/map format and
SAMtools. Bioinformatics 25, 2078-2079.
McKenna, A., Hanna, M., Banks, E., Sivachenko, A., Cibulskis, K., Kernytsky, A.,
Garimella, K., Altshuler, D., Gabriel, S., Daly, M. et al. (2010). The Genome
Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA
sequencing data. Genome Res. 20, 1297-1303.
Miller, A. C., Obholzer, N. D., Shah, A. N., Megason, S. G. and Moens, C. B.
(2013). RNA-seq-based mapping and candidate identification of mutations from
forward genetic screens. Genome Res. 23, 679-686.
Miskinyte, S., Butler, M. G., Hervé , D., Sarret, C., Nicolino, M., Petralia, J. D.,
Bergametti, F., Arnould, M., Pham, V. N., Gore, A. V. et al. (2011). Loss of
BRCC3 deubiquitinating enzyme leads to abnormal angiogenesis and is
associated with syndromic moyamoya. Am. J. Hum. Genet. 88, 718-728.
Nasevicius, A. and Ekker, S. C. (2000). Effective targeted gene ‘knockdown’ in
zebrafish. Nat. Genet. 26, 216-220.
Nechiporuk, A., Finney, J. E., Keating, M. T. and Johnson, S. L. (1999).
Assessment of polymorphism in zebrafish mapping strains. Genome Res. 9,
1231-1238.
Obholzer, N., Swinburne, I. A., Schwab, E., Nechiporuk, A. V., Nicolson, T. and
Megason, S. G. (2012). Rapid positional cloning of zebrafish mutations by linkage
and homozygosity mapping using whole-genome sequencing. Development 139,
4280-4290.
Patowary, A., Purkanti, R., Singh, M., Chauhan, R., Singh, A. R., Swarnkar, M.,
Singh, N., Pandey, V., Torroja, C., Clark, M. D. et al. (2013). A sequence-based
variation map of zebrafish. Zebrafish 10, 15-20.
Patton, E. E. and Zon, L. I. (2001). The art and design of genetic screens: zebrafish.
Nat. Rev. Genet. 2, 956-966.
Quinlan, A. R. and Hall, I. M. (2010). BEDTools: a flexible suite of utilities for
comparing genomic features. Bioinformatics 26, 841-842.
Rauch, G.-J., Granato, M. and Haffter, P. (1997). A polymorphic zebrafish line for
genetic mapping using SSLPs on high-percentage agarose gels. In Technical
Tips Online: BioMedNet.
Rozen, S. and Skaletsky, H. (2000). Primer3 on the WWW for general users and for
biologist programmers. Methods Mol. Biol. 132, 365-386.
Sander, J. D., Cade, L., Khayter, C., Reyon, D., Peterson, R. T., Joung, J. K. and
Yeh, J.-R. J. (2011). Targeted gene disruption in somatic zebrafish cells using
engineered TALENs. Nat. Biotechnol. 29, 697-698.
Solnica-Krezel, L., Schier, A. F. and Driever, W. (1994). Efficient recovery of ENUinduced mutations from the zebrafish germline. Genetics 136, 1401-1420.
1551
DEVELOPMENT
RESEARCH ARTICLE
RESEARCH ARTICLE
mapping and identification of a zebrafish ENU-induced mutation by wholegenome sequencing. PLoS ONE 7, e34671.
Wang, K., Li, M. and Hakonarson, H. (2010). ANNOVAR: functional annotation of
genetic variants from high-throughput sequencing data. Nucleic Acids Res. 38, e164.
Wheeler, D. L., Barrett, T., Benson, D. A., Bryant, S. H., Canese, K., Chetvernin, V.,
Church, D. M., DiCuccio, M., Edgar, R., Federhen, S. et al. (2008). Database
resources of the National Center for Biotechnology Information. Nucleic Acids Res.
36 Suppl. 1, D13-D21.
Whiteley, A. R., Bhat, A., Martins, E. P., Mayden, R. L., Arunachalam, M., UusiHeikkila, S., Ahmed, A. T., Shrestha, J., Clark, M., Stemple, D. et al. (2011).
Population genomics of wild and laboratory zebrafish (Danio rerio). Mol. Ecol. 20,
4259-4276.
DEVELOPMENT
Stickney, H. L., Schmutz, J., Woods, I. G., Holtzer, C. C., Dickson, M. C., Kelly, P. D.,
Myers, R. M. and Talbot, W. S. (2002). Rapid mapping of zebrafish mutations with
SNPs and oligonucleotide microarrays. Genome Res. 12, 1929-1934.
Strauss, W. M. (2001). Preparation of genomic DNA from mammalian tissue. Curr.
Protoc. Mol. Biol. 42, 2.2.1-2.2.3.
Tresnak, I. (1981). The long-finned zebra Danio. Trop. Fish Hobby 29, 43-56.
Trevarrow, B. and Robison, B. (2004). Genetic backgrounds, standard lines, and
husbandry of zebrafish. Methods Cell Biol. 77, 599-616.
Turner, D. J., Keane, T. M., Sudbery, I. and Adams, D. J. (2009). Next-generation
sequencing of vertebrate experimental organisms. Mamm. Genome 20, 327-338.
Voz, M. L., Coppieters, W., Manfroid, I., Baudhuin, A., Von Berg, V., Charlier, C.,
Meyer, D., Driever, W., Martial, J. A. and Peers, B. (2012). Fast homozygosity
Development (2015) 142, 1542-1552 doi:10.1242/dev.118786
1552