regardless of whether we used a short- or long-
lived species reference (c^2 ,P<1×10−^15 ). The
greatest cluster of positive associations was on
chromosome 1 (Fig. 3A); it extended for almost
250 kb and overlapped nine genes (Fig. 3B)
ranging from zero to seven copies among
different species (Fig. 3C). These nine genes
included six members of the butyrophilin
gene family (BTNandBTNLgenes) (Fig. 3D), a
family of immune regulators associated withhuman inflammatory diseases ( 22 ). Using our
linear model, we found that this signature is
primarily driven by the depth-associated com-
ponent of life span (Fig. 2H and fig. S15). Mu-
tations inBTNL2cause sarcoidosis in humans844 12 NOVEMBER 2021¥VOL 374 ISSUE 6569 science.orgSCIENCE
DSB repair via HR
DNA rec repair
DSB repair0246Short
LivedLong
Livedlog10Pq<0.05A B- 105 years
S. ruberrimus - 147 years- 116 years
- 105 years
- 205 years
- 118 years
long-livedlong-livedshort-livedshort-livedlong-livedlong-livedshort-livedshort-livedCDR
2
0.59
05010015020050 100 150
predicted lifespan (L~S+D)lifespanE F010203040Residual
Maximum DepthSize at Maturity% explainedADRB3GCGRMLXIPLAKT1FOXO1IGF1G−0.6 −0.3 0.0 0.325507510025507510
0
25 50 75 100residualsize at maturitymaximum depth(R)esidual(S)ize (D)epthlog 10 P
4
8
12
16SH
TP53I3VWFCYP8B1MLECADRB3DNAJB2NRG1NEK6PUM2BRIP1 LACC1PNLIPARFGAP1NUDT5 AQP1GCN1RDTRAF3IP3
TNFRSF6B
MADD
NECTIN3 RPL23RS DInsulin Signalling (72)RS DRibosome (49)
RS DGlycerolipid metabolism (24)RS DRS DResponse to Hypoxia (45)
RS DTelomere Maintenance (28)
RS DAcute Myeloid Leukemia (28)RS DHeat Shock Protein Binding (62)
RS DAntigen Processing & Presentation (21)
RS DApoptosis (41)I−3
030 200 400600 821
MCM6−3
030 200
RAD51AP1
−3
030 200 400 600754
NBN−3
030
SWI5−3
0
30 200 400
DCLRE1B−3^030 200
FEN1−3^030 200 400 600756
SPIRE1−3^030 200 400502
WAS
−3
0
30 200 400600 821
MCM6−3
0
30 1000 2000 2804
NIPBL
−3
030 1017
FAN1−3
030 200 400 548−3
030 200 400 633
XRCC1−3
030 200 400 600754
NBN
−3
030 200
FEN1
−3^030 1162
KDM2A−3
030 1000
SLX4−3
0
30 1000
SLX4−3
030 200 400 600777
BARD1−3
030 200400600 897
RBBP8
−3
030 200400600 897
RBBP8Fig. 2. Genetic underpinnings of life-span adaptations.(A) Distribution of
pathway enrichmentPvalues for genes under positive selection in short- and
long-lived species. DSB, double-strand break; HR, homologous recombination; rec,
recombination. (B) Schematics of the 16 DNA replication, repair, or maintenance
genes exhibiting positive selection in long-lived rockfishes. Domain structure
(bottom), gray lines indicate putative functional consequence of rockfish amino
acid compared with human (PROVEAN score). Red and green bars indicate
mutations distinct from all non-long-lived taxa colored by SIFT score−tolerated
(green) / damaging (red). (C) Schematic representation of relative evolutionary
rate test and (D)Pvalue distribution of genes from relative evolutionary
rate test (RERconverge) of correlation between evolutionary rate and individual
traits. Life span and the individual components of a predictive linear model of
life span (E) are used as traits. Gene names are highlighted for the top 12 genes
associated with the linear model residual as a trait in addition to four other
significant gene candidates previously associated with aging (full gene list in
table S14). (E) Linear regression model between maximum life span and
predicted life span based on body size and depth for different species.
(F) Proportion of variation in life span explained by size, depth, and the residual
from the linear model. (G) Relative evolutionary rates of six genes involved
in glucose, insulin, and nutrient signaling, four of which have been associated
with aging in other taxa (table S14). (H) Ternary plot of 91 life spanÐassociated
genes plotted as a function of the relative importance along linear model
axes, with dashed line at 50% (inset with gene labels) and (I) heatmaps of
pathway enrichment along axes.RESEARCH | RESEARCH ARTICLES