Genetic analysis of cavefish reveals molecular convergence
in the evolution of albinism
Meredith E Protas
1
, Candace Hersey
2
, Dawn Kochanek
3
, Yi Zhou
2
, Horst Wilkens
4
, William R Jeffery
5
,
Leonard I Zon
2
, Richard Borowsky
3
& Clifford J Tabin
1
The genetic basis of vertebrate morphological evolution has
traditionally been very difficult to examine in naturally
occurring populations. Here we describe the generation of a
genome-wide linkage map to allow quantitative trait analysis of
evolutionarily derived morphologies in the Mexican cave tetra,
a species that has, in a series of independent caves, repeatedly
evolved specialized characteristics adapted to a unique and
well-studied ecological environment. We focused on the trait
of albinism and discovered that it is linked to Oca2,aknown
pigmentation gene, in two cave populations. We found
different deletions in Oca2 in each population and, using
a cell-based assay, showed that both cause loss of function
of the corresponding protein, OCA2. Thus, the two cave
populations evolved albinism independently, through
similar mutational events.
The relatively closed, often nutrient-poor, and lightless environment
of caves represents a marked change in ecological conditions to which
several entrapped species have adapted. Obligate cave-dwelling ani-
mals, called troglobites or troglodytes, are characterized by a remark-
able convergence of eye and pigment loss across diverse species such as
spiders, isopods, salamanders and fish
1
.
There are 86 known troglodytic species of fish
2
. The best studied is
the Mexican tetra, identified by some authors as Astyanax mexicanus
and others as Astyanax fasciatus; the two names should be considered
synonymous in the present context and the species will be referred to
herein as Astyanax. This species has 29 cave populations in the karst
region of the Sierra de El Abra of northeast Mexico and one additional
population in Guerrero (Fig. 1a)
3,4
. A surface, or river-dwelling, sister
population of the cave morph lives in southern Texas and northeastern
Mexico and can still interbreed with the cave morph. Phenotypically,
the cave and surface morphs are very different; among other char-
acteristics, the cave morph has a greater weight per unit length, less
pigment, regressed eyes, larger nostrils, more maxillary teeth, more
cranial neuromasts and more taste buds, as well as differences in
feeding, schooling and aggressive behaviors (Fig. 1b–d)
4,5
.Molecular
phylogenetic studies indicate that several cave populations indepen-
dently evolved these characteristics
6–8
.
To provide a framework in which to study the genetics of this
species, we made a microsatellite linkage map. We have isolated and
Rio Sabinas
Rio Frio
Molino
a b
c
d
Pachón
Pachón
Japonés
Ciudad Valles
kilometers
0 5 10 15
Ciudad Mante
N
Surface
Molino
R
i
o
M
a
n
t
e
A
r
r
o
y
o
L
a
g
a
r
t
o
S
I
E
R
R
A
D
E
E
L
A
B
R
A
S
I
E
R
R
A
D
E
C
O
L
M
E
N
A
R
io
V
a
l
l
e
s
Rio
T
a
m
p
a
a
n
M
E
X
I
C
O
Figure 1 Phenotype and locations of albino cave populations of Astyanax
mexicanus.(a) Map of the area in Mexico where the different cave
populations are found. Dots represent cave populations. Caves with red dots
are Molino, Pacho´n and Japone´s, all of which contain a majority of albino
individuals. Inset map at bottom shows the location of the region within
Mexico. (b) A representative surface fish. (c) A representative Molino
cavefish. (d) A representative Pacho´n cavefish.
Received 20 July; accepted 13 October; published online 11 December 2005; doi:10.1038/ng1700
1
Department of Genetics, Harvard Medical School, Boston, Massachusetts 02115, USA.
2
Children’s Hospital Stem Cell Program, Department of Hematology/Oncology,
Howard Hughes Medical Institute, Children’s Hospital Boston, Boston, Massachusetts 02115, USA.
3
Cave Biology Research Group, Department of Biology, New York
University, 1009 Main, 100 Washington Square East, New York, New York 10003, USA.
4
Zoological Institute and Zoological Museum, University of Hamburg, Martin-
Luther-King-Platz 3, 20146 Hamburg, Germany.
5
Department of Biology, University of Maryland, College Park, Maryland 20742, USA. Correspondence should be
addressed to C.J.T. (tabin@genetics.med.harvard.edu).
NATURE GENETICS VOLUME 38
[
NUMBER 1
[
JANUARY 2006 107
LETTERS
©
2
0
0
6

N
a
t
u
r
e

P
u
b
l
i
s
h
i
n
g

G
r
o
u
p


h
t
t
p
:
/
/
w
w
w
.
n
a
t
u
r
e
.
c
o
m
/
n
a
t
u
r
e
g
e
n
e
t
i
c
s

designed primers corresponding to over 600 microsatellites. Using a
backcross from the Molino cave population with 111 progeny, we
obtained 35 linkage groups composed of 267 markers, out of 300
markers genotyped, with a coverage of 1,916 cM (Fig. 2 and
Supplementary Table 1 online). Astyanax has 25 chromosomes, sug-
gesting that with the addition of more markers, some of the linkage
groups would collapse. We are also genotyping a larger F
2
cross from
another cave, Pacho´ n, which should coalesce some of the linkage
groups and allow for comparisons between the two cave populations.
We identified a number of statistically significant, quantitative loci
for different traits present in the cave form, most of which will be
described elsewhere. Here we focus on one such trait, albinism.
Previous genetic studies have indicated that albinism in the Pacho´ n
cave is caused by a single recessive mutation
9,10
.IntheMolino
backcross, albinism mapped to a single locus in linkage group 16
with a LOD score of 17.29 at microsatellite marker 218E, accounting
for 49.4% of the variance in this trait (Fig. 3a). A similar analysis of
the Pacho´ nF
2
cross mapped the locus for albinism to the same
location with a LOD score of 17.98 at marker 218E, accounting for
42.6% of the variance in this trait (data not shown). This coincidence
of loci responsible for albinism raises the following three possibilities:
the two cave populations could have the same mutation in the same
gene, different mutations in the same gene or mutations in distinct
but closely linked genes. To address the latter possibility, we performed
a complementation test between a Molino individual and a Pacho´ n
individual, which yielded only albino offspring (Fig. 3b). Thus,
albinism in these two cave populations is caused by mutations in
the same gene.
To identify the gene responsible for albinism in Astyanax,we
genotyped individuals of the Molino backcross for a series of candi-
date genes, based on known albinism loci in mouse and humans:
tyrosinase (Ty r), tyrosinase-related protein-1 (Tyrp1) and ocular and
cutaneous albinism-2 (Oca2)(Fig. 2). One of these genes, Oca2,
mapped to the albino locus and increased the LOD score in the
Molino backcross to 68.6 (Fig. 3c), now accounting for 93.1% of the
variance of this trait, and the LOD score in the Pacho´ nF
2
cross to
60.66, now accounting for 71.6% of the variance of this trait (data
not shown). Furthermore, there is a perfect association between
the genotype of the Oca2 marker and the phenotype of albinism in
all successfully genotyped individuals of both the Molino backcross
(105 individuals) and the Pacho´ nF
2
cross (215 individuals).
Although the function of Oca2 is unknown, it is the most commonly
mutated gene in cases of human albinism
11
and is also responsible for
pigmentation phenotypes in mouse and medaka
12,13
. To test whether
Oca2 mutations are responsible for albinism in cavefish, and, more
importantly, to identify the specific genetic lesions in Oca2 responsible
for albinism, we compared the sequence of the Oca2 cDNA in surface,
Pacho´n and Molino individuals (Supplementary Fig. 1 online). We
found numerous differences in the Oca2 sequences present in the two
cave populations as compared to their surface counterparts (Fig. 4a
and Supplementary Fig. 1). The Pacho´ n cave population had three
polymorphisms that could affect Oca2 function: two were amino acid
changes in conserved residues, and the last was a deletion extending
from within intron 23 through most of exon 24, such that the cDNA
includes part of intron 23 fused to the last nine base pairs of exon 24
and the 3¢ UTR (Fig. 4a and Supplementary Fig. 1). The Molino cave
population had only one major difference: exon 21 was missing
(Fig. 4a and Supplementary Fig. 1).
The missing exons observed in the Molino and Pacho´ n Oca2
sequence could, in principle, be explained by either alterations in
splicing or deletions of genomic DNA. Amplifying from genomic
DNA, we found that in both cases, the observed losses of exonic
sequence in the Molino and Pacho´ n Oca2 cDNAs were attributable to
genomic deletions (Supplementary Fig. 2 online).
24 206A
209E0
18
6
0
103A
219A
207F
11
6
0
15A
232A
231C
11
0
129B
43A
32 33 34 35
113 33b
95
90
213E
237A
213F82
80
73
66 30C
26A
105B
43C54
50
44
218E
132B
34
205E
210A
9
0
88
81
75 17A
220B
128B
205D61
9C
208E
57
53
44
37
29
221A
142A
28A
236A
7B0
8
0
7
12
113B
217A
23A
28
34
112A
221B
36D46
53
74 210D
132C
88
89
99 202B
224A
224C
123A64
35
32
23
19
16
10
7
0
216D
116C
6C
214E
226E
229D
214C
230A
0
8
10
20
37
59
63
68
73 55b
218A
207C
144A
19B
104C
203E
131A
55a 0 207D
18 227A
222E23
37 208A
3A41
47 109C
119A52
66 113C
139B23
12
10
6
0
207A
228E227D
103C
203A
47 115C
15C35
30
27
19
228D
42A
139A
220E0
42
40
31
21
0 219B
111B
109B
219E
226B
49 105A
223C
229B
21
16
7
0
235D
119B
58 3c
47 3B
127B35
16 8C
133C0
51
43
37 17C
214A
211E
143A0 0
16
24 232C
207E234C
220A
38
30
18
0 230E
5C
132A
237B
0
3
6
10 210B
216E
237F
230C
22
9
0
225A
239e
204D
31302928272625242322212019181716
Oca2
Tyrp1
Tyr
0 237E
0
0
00 0
0
0
00 0
0
0 00123C
135C
211B
235B
241B
224F
217E
240E
202A
208C
24C
104B
114A
226A
242A
232D
229A
205A
214D
13
15
27
53
59
63
69
71
72
74
81
15
17
19
21
34
47
73
1B
135A
222C
130A
142B
122B
207B
215F
215B
220D
224E
13
15
21
35
38
40
44
54
56
59
67
229C
11B
203D
136A
1
14
22
4
16
31
54
68
74
77
79
83
98
107
118
127
131
137
143
144
146
151
6
204F 111C 118C
222B
114B
211D
127A
106B
239a
102C
206E
44A
41C
1C
215D
224B
241A
239c
227B
129A
19C
16B
211A
236B
214B
11
18
235E
117C
112B
217B
6A2
3
14
19
226F
234A
241D
206D
208D
107C
213A
235A
231A
235F
237C 228C
6
7
9
13
17
20
22
39
41
45
240B
119C
103B
110B
209A
238A
218D
44C
203F
234B
108C
143C
120C
216F
17B
125C
142C
204E
122C
120B
33c
222D 216B
5
16
32
38
39
47
71
76
96
101
103
111
114
119
125
144
154
163
166
168
172
213D 111A
225B
39C
240A
145A
217D
209D
27C
2B
23B
216C
214F
242B 239d9
10
19
32
38
12
15
22
53
68
91
30B
101A
106A
26C
107B
230D
101B
123B
120A
229E
114C
216A
7
16
20
21
23
25
28
29
30
32
35
218B
37C
139C
215C
16A
24B
205F
213B
202D
241C
202E
133B
209B
44B
225C
226D
128A
209C
211C
136B
16C
39a
1
11
28
31
44
47
52
55
56
80
83
96
99
109
125
130
138
142
151
156
171
182
12 3 45 6 7 89101 12131415
Figure 2 Microsatellite linkage map of Astyanax mexicanus. The names of the microsatellite markers are to the right of each linkage group and the positions
of the markers (in cM) are on the left. Markers in three genes (Oca2, Tyr and Tyrp1) are labeled in red.
108 VOLUME 38
[
NUMBER 1
[
JANUARY 2006 NATURE GENETICS
LETTERS
©
2
0
0
6

N
a
t
u
r
e

P
u
b
l
i
s
h
i
n
g

G
r
o
u
p


h
t
t
p
:
/
/
w
w
w
.
n
a
t
u
r
e
.
c
o
m
/
n
a
t
u
r
e
g
e
n
e
t
i
c
s