Journal of Gerontology: BIOLOGICAL SCIENCES Copyright 2000 by The Gerontological Society of America

2000, Vol. 55A, No. 9, B455–B461

Differential Longevity in Mouse Stocks Selected for

Early Life Growth Trajectory
Richard A. Miller,1 Clarence Chrisp,2 and William Atchley3

1Department of Pathology, University of Michigan School of Medicine; Institute of Gerontology, University of Michigan;

Downloaded from http://biomedgerontology.oxfordjournals.org/ at Rutgers University Libraries/Technical Services on April 10, 2015
and Ann Arbor DVA Medical Center.
2Unit for Laboratory Animal Medicine, University of Michigan School of Medicine, Ann Arbor.
3Department of Genetics, North Carolina State University, Raleigh.

Small body size is associated with superior longevity in several intraspecies comparisons, including dogs bred for
specific forms of work, mice and rats fed diets low in calories, rats fed diets low in methionine, and mutant mice
whose levels of growth hormone and thyroid hormone are atypically low. To further investigate the interactions
among body size, genetic endowment, and longevity, we measured the life span of female mice selectively bred
from Institute for Cancer Research stock for differences in rate of body weight gain. These mice were selected for
differential rates of growth either early (0–10 days) or later (26–56 days) in the first 2 months of life. The data
show a good correlation between the average weight of the stock and its mean longevity, with low body size asso-
ciated, as predicted, with longer life span. Weight at 3, 6, and 12 months, and weight at peak body weight, are all
significant predictors of longevity (among stocks) in univariate regressions; weight at 6 months has the strongest
association in stepwise multiple regression. There is no significant correlation between the life span for the stock
and the proportion of deaths attributable to neoplasia in this group of mice. The data provide support for the hy-
pothesis that genetic factors that influence early life growth trajectories can have a strong influence on life span.
These size-selected mice provide useful tools for analysis of the genetic factors that influence life history parame-
ters, including maturation and aging rates.

Three others (LB ⫽ “late big”) were selected for high weight
S OME authors have suggested that genetic factors influ-
encing longevity and disease risk in later life represent
pleiotropic effects of polymorphic alleles selected to im-
gain from 28–56 days, combined with average gain from day
of birth to day 10. Three stocks (ES = “early small”) were se-
prove reproductive fitness earlier in life (1–3). Despite con- lected for exceptionally slow weight gain from day of birth to
siderable useful speculation, there is little direct evidence day 10, with average gain at 28–56 days. Three stocks (LS)
about the number and nature of loci whose effects on early were selected for slow gain from day 28–56, and average gain
development and fitness actually influence longevity and at 0–10 days. Lastly, three control stocks (C) were bred in par-
late life disease risk. allel, from the same initial pool of heterogeneous progenitors,
Several instances have been reported where differences without deliberate selection for weight gain trajectories.
among individuals in body size are associated with differences This report presents life span and pathology data for mice
in longevity within species. In each case, superior longevity is in these 15 size-selected mouse populations.
associated with smaller body size. In some cases the differ-
ences in body size are a result of dietary manipulation, as in MATERIALS AND METHODS
studies of caloric restriction in mice and rats (4) or lifelong
methionine deprivation in rats (5). In other circumstances the Mouse Selection
differences in body size reflect differences in single genetic Foundation of these selection lines derives from 180 litters
loci, as in the df/df dwarf mice (6) and the urokinase knockout of ICR (Institute for Cancer Research, London, England)
mice (7). Differences in body size, associated with altered lon- mice obtained in 1986 from Sprague-Dawley (Indianapolis,
gevity, can also be produced by natural selection: among dogs, IN). These litters, produced from 261 female and 256 male
for example, differences among breeds in body weight are founders, were randomly allocated into three replicates for
very strongly correlated to interbreed differences in mean life each selection and control strain, producing a total of 15
span (8), and similar correlations have been noted in popula- lines. Each replicate contained 12 litters. At each generation
tions of flies (9), whose size differences represent selected ge- mice were selected according to a restricted selection index
netic adaptations to specific ecological niches. procedure that ensures that growth at later intervals is held
To seek further insights into the relationship between body constant when selection is focused on early rate of develop-
size and life span, we have examined longevity in a series of ment in body weight, and vice versa (10). Within-family
15 mouse stocks that have been selected over 22 generations selection was practiced to minimize maternal effects. Litters
for differential rate of body weight by restricted index selec- were standardized at birth to eight pups and an equal sex
tion (10). Three of the stocks (EB ⫽ “early big”) were se- ratio. Litters with fewer than eight pups were augmented
lected for rapid gain in body weight from birth to 10 days of with excess pups from other litters. These excess pups were
life, combined with average weight gain from 28–56 days. tail-clipped to distinguish them from the original members

B455
B456 MILLER ET AL.

of the litter. The substitute mice did not enter into any calcu- later tested for antiviral antibodies and for parasites; all such
lations for experiments. All mice were forcibly weaned at tests were negative throughout the course of this study.
21 days of age. Mice found dead were subjected to a careful gross
Two selection treatments (E) focused on increased or de- necropsy, and to a detailed histopathological examination
creased rate of development in body weight from birth to 10 using methods that have been described in detail elsewhere
days of age. During this interval, postnatal growth is most (15).
influenced by changes in cell number (hyperplasia) in rele-
vant organs and tissues (11–13) (and Atchley et al., unpub- RESULTS
lished data). The other two selection treatments (L) focused
on increased or decreased rate of development in body Relation of life span to interstock differences in body

Downloaded from http://biomedgerontology.oxfordjournals.org/ at Rutgers University Libraries/Technical Services on April 10, 2015
weight between 28 and 56 days of age. During this interval, weight.—Table 1 provides the mean (⫾SD), median, and
growth is more influenced by changes in cell size (Atchley maximal observed longevity, as well as mean weight at 6
et al., unpublished data). months and the peak weight for each of the 15 stocks. The
The three replicated control lines were randomly selected sample sizes used to calculate the mean weights are slightly
using a random number generator to produce a pseudo se- lower than those used in the longevity calculations because
lection index score. Cellular analyses of these mouse stocks three mice received as weanlings died before the first
indicate that for several organ systems, selection for rate of weight determination at 6 months of age. The 15 stocks are
body weight gain between birth and 10 days of age changed listed in descending order ranked by mean life span.
the number of cells (hyperplasia), whereas selection for rate There has been significant direct and correlated response
of gain between 28 and 56 days of age altered cell size (hy- to selection for differential early and late rates of develop-
pertrophy) (Wei et al., unpublished data). ment in these mice. For example, there has been a 2.7-fold
The three replicated stocks designated here as EB (early range (from 344 to 941 days) in mean stock life span, and a
big) were those selected for rapid weight gain between birth 2.0-fold range in peak body weight (33 to 65 grams). Thus,
and 10 days of age while holding late gain constant. The while these stocks were derived from a common random-
three replicated LB (late big) strains were selected for high bred ICR stock of progenitor mice, the various selection
weight gain from 28–56 days holding gain between birth lines have undergone significant genetic divergence in a
and 10 days of age constant. Three replicated ES stocks number of characteristics.
(early small) were selected for decreased rate of weight gain A protected least significant difference test (16) was car-
from day 0–10 holding 28–56 day gain constant and three ried out to partition the mouse strains into statistically ho-
replicated LS (late small) stocks were selected for decreased mogeneous subsets for each trait. For 6 months body
gain from day 28–56 holding 0–10 day gain constant. weight, there are five homogeneous subsets. The homoge-
Previous data (10) have shown that various EB and LB neous subset containing the five heaviest mouse strains in-
replicates had statistically indistinguishable mean body cludes strains LB3, LB1, LB2, EB2, and EB3. Thus, those
weights, tail lengths, and organ weights at 56 and 91 days of mice selected for gain in body weight from 28–56 days
age. These authors described this phenomenon as develop- are the heaviest at 6 months of age, followed by two of
mental homoplasy (structural resemblance due to convergent the strains selected for gain from birth to 10 days of age.
evolution rather than common ancestry) to reflect that the For average longevity, there are five statistically homoge-
same complex morphological phenotype can be produced by neous groups including (from longest to shortest lived): (i)
quite different genetic and developmental processes. The im- LS2, EB1, and C2; (ii) EB1, C2, LS1, C1, LS3, EB3, and
pact of selection in these mice on maternal effects, organ size, LB2; (iii) LS1, C1, LS3, EB3, LB2, and ES3; (iv) LS3,
and other attributes is described separately (14) and in a series EB3, LB2, ES3, ES2, LB1, ES1, and D3; and (v) C3, EB2
of submitted manuscripts (Ernst et al., 1999a, 1999b; Cren- and LB3.
shaw et al., 1999; Atchley et al., 1999). Figure 1 shows a scatterplot relating mean life span for
each stock to mean body weight at 6 months of age. In this
Method plot, the area of the circle is proportional to the number of
For the present longevity experiment, weaned female mice individual mice tested from each indicated stock. Using the
were produced from the selection colony at North Carolina means of each stock, univariate linear regression analysis
State University and shipped at about 25 days of age to the shows that weight at age 6 months is a good predictor of
University of Michigan. At Michigan the mice were housed mean life span of the stock (r ⫽ .69, p ⫽ .004). This plot
in a specific pathogen-free colony and observed at least suggests that low body weight is associated with increased
daily until their natural death. Mice were housed initially at life span. Maximum observed life span per stock is also
3–4 mice/cage and given free access to Purina Mouse Chow strongly correlated with weight at age 6 months (r ⫽ .74,
and tap water. All mice were housed in the same room p ⫽ .002; scatterplot not shown), although the value of this
throughout the study period. Mice were weighed once a result is seriously compromised by the small numbers of
month beginning at 6 months of age. Each cage was pro- mice studied in each stock. Weight at age 12 months (r ⫽
tected by a filter-paper bonnet to minimize the risks of air- .57, p ⫽ .03) and lifetime peak weight (r ⫽ .61, p ⫽ .02) are
borne infection. The pathogen-free status of the colony was associated with stock mean life span almost as strongly as
documented every 3 months using a procedure in which the weight at 6 months, and indeed among the 15 stocks
spent bedding from the experimental mice was provided to these three estimates of stock size are strongly correlated,
sentinel mice of the outbred CD1 stock, and the sentinels with Pearson r correlations between .91 and .97.
 BODY SIZE AND LONGEVITY IN MICE B457

Table 1. Life Span and Weight Statistics for 15 Mouse Stocks

Life-Span Weight Weight-6 Rank Weight Peak Wt
Stock Mean LS n Cluster Median Min Max (6 mon) n Cluster (a) (Peak) n Cluster
LS2 941 ⫾ 117 15 a 905 804 1146 29 ⫾ 2.4 15 ef 14 36 ⫾ 1.5 15 g
EB1 832 ⫾ 212 9 ab 899 304 1035 40 ⫾ 5.7 9 bc 7 52 ⫾ 7 9 de
C2 827 ⫾ 150 17 ab 879 568 1043 32 ⫾ 2.6 17 de 12 37 ⫾ 2.5 17 g
LS1 785 ⫾ 221 15 bc 825 150 1029 35 ⫾ 3.2 14 d 10 47 ⫾ 3.8 14 f
C1 777 ⫾ 139 19 bc 783 561 1136 34 ⫾ 3.0 19 d 11 43 ⫾ 2.9 19
LS3 719 ⫾ 250 12 bcd 802 362 1149 29 ⫾ 3.5 12 f 15 38 ⫾ 4.4 12 g
EB3 704 ⫾ 157 15 bcd 731 327 976 46 ⫾ 6.3 15 a 5 65 ⫾ 9.1 15 a

Downloaded from http://biomedgerontology.oxfordjournals.org/ at Rutgers University Libraries/Technical Services on April 10, 2015
LB2 693 ⫾ 147 6 bcd 700 521 854 49 ⫾ 2.8 6 a 2 53 ⫾ 2.3 6 ce
ES3 669 ⫾ 177 12 cd 687 209 888 32 ⫾ 2.9 11 def 13 33 ⫾ 10.5 12 g
ES2 664 ⫾ 187 21 d 627 321 1074 39 ⫾ 3.3 21 c 9 49 ⫾ 5.2 21 ef
LB1 615 ⫾ 211 14 d 663 244 889 49 ⫾ 4.9 14 a 3 58 ⫾ 5 14 bc
ES1 592 ⫾ 213 15 d 513 237 954 39 ⫾ 4.5 15 bc 8 51 ⫾ 5.4 15 ef
C3 572 ⫾ 168 5 de 593 299 759 43 ⫾ 4.8 5 b 6 57 ⫾ 6.2 5 bd
EB2 473 ⫾ 162 23 e 395 289 791 48 ⫾ 5.1 23 a 4 55 ⫾ 6.6 23 cd
LB3 344 ⫾ 114 4 e 363 188 462 51 ⫾ 12.7 3 a 1 61 ⫾ 4 3 ab

Note: Letters in the cluster columns indicate groups that are not statistically different from one another.

We also examined the association between stock life span dependent variable (life span in this case). At each step, a
(measured in this study) and measures of weight at 91 days new predictor variable is added based upon its partial corre-
based on mice at Generation 19 housed at North Carolina lation with the dependent variable. The goal is to produce a
State University (Crenshaw and Atchley et al., unpublished predictive equation that gives the best prediction of the de-
data). This weight measure, too, was strongly associated pendent variation using the smallest number of predictor
with mean stock life span (r ⫽ 0.59, p ⫽ 0.02), consistent variables.
with the interpretation that the associations noted above Table 2 gives the stepwise regression results and shows
were not dependent on any artifacts of shipping stress or id- that body weight at 6 months is the best predictor of stock
iosyncrasies of the husbandry procedures used at the Uni- life span (t ⫽ –2.40, p ⫽ .035). Body weight at 12 months is
versity of Michigan vivarium. positively correlated with stock life span, though at only
Body weights at various ages are not statistically inde- marginal significance ( p ⫽ .099), after the correlation with
pendent. Consequently, we performed a stepwise multiple body weight at 6 months is removed. Thus, although low
regression analysis, with mean stock life span as dependent body weight is associated with increased longevity, the abil-
variable and weight at 6, 12, and peak as potential indepen- ity to maintain or increase body weight between 6 and 12
dent variables. The stepwise regression procedure (16) adds months may also be associated with increased resistance to
body weight variables in terms of their ability to predict the late-life illness and early death.

Within selection treatments.—Some interesting trends

are evident within selection treatments (i.e., among repli-
cates for a given selection protocol). For example, among
the stocks selected for rapid gain between birth and 10 days,
EB2 and EB3 do not differ significantly in 6-month weight,
but EB1 had a significantly smaller body weight. In terms of
life span, EB1 was longer lived than EB2 and EB3. This dif-

Table 2. Stepwise Multiple Regression Analysis

Standard Standard
Error of Error
Beta* Beta b† of b t (11) p Level‡
Intercept 1149 159.4 7.21 .000
Weight, 6 months –1.29 .54 –25.6 10.7 –2.40 .035
Weight, 12 months 1.61 .89 31.5 17.5 1.80 .099
Peak weight –.99 .79 –15.6 12.4 –1.26 .235

Figure 1. Association of longevity with weight across mouse *Standardized regression coefficient.
†Nonstandardized regression coefficient for equation in which stock life span
stocks. Each symbol represents the mean weight at 6 months for mice
of the indicated stock and the mean longevity of mice of that stock. is the dependent variable and independent variables are weight (at 6 months),
The area of each symbol is proportional to the number of mice tested, weight (at 12 months), and peak weight.
‡Significance level for the regression coefficient for each independent variable;
which varies from 4 mice (stock LB3) to 23 mice (stock EB2). There
is a significant correlation between these measures (r ⫽ .69, p ⫽ .004). only weight (at 6 months) is significant at p ⬍ .05.
B458 MILLER ET AL.

ference in longevity is not statistically significant, but this longevity for the 13 stocks with N ⬎ 5 mice showed a
lack of significance could be due to the small sample size in strong association between inbreeding coefficient and lon-
EB1 (n ⫽ 9). gevity among 11 of the stocks, with two outliers (EB2 and
In the three control replicates (C1, C2 and C3), random ES1). This association between high longevity and low in-
mating occurred within each replicate stock. The small pop- breeding coefficient, which did not reach statistical signifi-
ulation size (12 pairs per generation) might be expected to cance in this group of stocks, is consistent with many previ-
lead to stochastic variation in trait means despite the ab- ous analyses of the impact of inbreeding on longevity.
sence of directional selection. Indeed, there are significant To see if the association between stock weight and lon-
differences in 6-month weight in these three lines, with C3 gevity was influenced by interstock differences in inbreed-
being significantly heavier than C1 and C2 (these latter two ing, we calculated a multiple regression using weight (at age

Downloaded from http://biomedgerontology.oxfordjournals.org/ at Rutgers University Libraries/Technical Services on April 10, 2015
strains being not significantly different for the trait). With 6 months) and inbreeding coefficient as independent vari-
regard to longevity, C1 and C2 do not differ significantly ables and stock life span as the dependent variable. The ef-
(C1 ⫽ 777 ⫾ 139 days and C2 ⫽ 827 ⫾ 150 days). C3, on fect of weight in this regression equation was highly signifi-
the other hand, had a mean longevity of 572 ⫾ 168 days), cant (r ⫽ .70, p ⫽ .01), but there was no significant effect of
which was significantly less than seen in C1 and C2. Thus, inbreeding coefficient (p ⫽ .62). We conclude that the
within the randomly selected control replicates, as in the set weight/longevity association does not reflect an unsus-
of EB stocks, there seems to be an inverse relationship be- pected influence of inbreeding on these two factors.
tween body weight at 6 months of age and longevity (i.e.,
the significantly heavier C3 replicate has significantly Necropsy results.—Of the 202 mice originally entered
shorter average life span). into this study as weanlings, 181 were discovered soon
enough after death to make a necropsy potentially informa-
Within genetic strain analyses.—Our current data set tive. Of these 181 cases, 17 (9%) were uninformative, in
provides only limited statistical power for testing the hy- that no conclusion could be drawn as to likely cause of
pothesis that body weight is associated with longevity death, usually due to advanced autolysis. In eight of the re-
among individual mice within strains. This lack of power maining cases, death was attributed to two serious diseases,
stems from small sample sizes, as our test group contained one neoplastic and one not neoplastic, under such circum-
only 3–23 evaluable mice in any one stock. Among the nine stances that it was not possible to assign either disease, by
stocks where N ⱖ 14, there were no significant correlations itself, as the likely cause of death.
between mouse longevity and mouse weight at 6 months, A synopsis of the remaining 156 diagnoses is presented
nor any stocks in which r ⬎ .50; among these nine stocks, in Table 3. Among the 156 diagnosable cases, 5 different di-
the slope of regression line was negative in six cases. agnoses each accounted for at least 5% of the deaths: lym-
We also sought evidence of a correlation between indi- phoma (29 cases), pituitary adenoma (19 cases, of which 16
vidual mouse life span and body weight. As expected, a re- were found in EB2 mice), congestive heart failure (15
gression between mouse weight at 6 months and mouse life cases), fibrosarcoma (11 cases), and pulmonary adenocarci-
span was highly significant (r ⫽ –.42, p ⬍ .0001, N ⫽ 199), noma (10 cases). Thirty-six mice (23%) died of non-neo-
reflecting the strong interstock differences in both life span plastic lesions, in addition to the 15 that died of congestive
and weight. When individual mouse weights were normal- heart failure; these lesions included cases of glomerulone-
ized, however (i.e., expressed as the number of standard de- phritis (n ⫽ 6), ovarian hematoma (n ⫽ 4), and dermatitis (n ⫽
viations above or below the mean weight for the given 4), in addition to a variety of less common inflammatory
stock), the relationship between normalized weight and lon- conditions (pancreatitis, osteomyelitis, peritonitis, pneumo-
gevity was small and not statistically significant (r ⫽ .05, nia with sepsis, etc.), and five cases in which combinations
p ⫽ .45). Thus, within the limits imposed by our population of non-neoplastic diseases contributed to death.
sizes, the factors that influence the strong correlation among Among the 156 diagnoses, neoplasia was responsible for
stocks between body weight and life span do not seem to 105 deaths (67%), but the proportion of deaths due to neo-
have appreciable effect among mice within these stocks. plasia varied among stocks from 20% (1 of 5 cases in LB2)
to 86% (6 of 7 deaths in EB1). Averaged across the 15
Impact of inbreeding on longevity.—Even with optimal stocks, the mean proportion of deaths attributable to neopla-
breeding methods, any long-term selection experiment sia was 63%. We used the proportions test of S-Plus (Math-
inevitably increases the level of inbreeding among the se- Soft, Inc., Seattle, WA) to test the null hypothesis that the
lected and control stocks. The level of inbreeding may influ- proportion of deaths due to neoplastic lesions did not differ
ence longevity because of increased levels of homozygos- among the 15 stocks, and obtained a marginal p ⫽ .17. We
ity, the accumulation of alleles of reduced fitness, and therefore cannot conclude that these stocks differed among
related effects. We have computed estimates of the levels of one another in this trait. A plot of mean longevity versus the
inbreeding in females in these selection lines using the IN- proportion of deaths due to neoplasia revealed no evidence
BREED Procedure in the SAS Statistical Package (SAS of an association between these two measures.
Technical Report P-229, SAS/STAT Software: Changes The number of deaths due to any single cause in any one
and Enhancements, Release 6.07, Cary, NC). These values stock was in most cases too low to provide much statistical
range from 0.232 in LS2 to 0.363 in LB1. A univariate re- power for testing hypotheses of stock-specific mortality pat-
gression of life span on inbreeding coefficient yielded r ⫽ terns. The two exceptions are shown in Table 3. Among the
.33 ( p ⫽ .23). A scatterplot of inbreeding coefficient versus 19 cases for stock EB2 mice, 16 deaths were attributable to
 BODY SIZE AND LONGEVITY IN MICE B459

Table 3. Necropsy Results least partly by common influences within a species, by pro-
%
viding another example where small size is associated with
% Pituitary % superior longevity.
Stock Cases* Diagnoses† Neoplastic‡ Adenoma Lymphoma Many reports have associated longer life span with smaller
C1 19 17 82% 0% 18%
body size. The best known and most often replicated associ-
C2 14 13 54 0 8 ation comes from an environmental intervention, in which
C3 5 4 50 0 0 caloric deprivation leads to dramatic life span extension in
EB1 9 7 86 0 14 mice and rats (4). Although the effect varies somewhat with
EB2 21 19 84 84 0 strain and dietary protocol, mice or rats allowed to eat only
EB3 14 11 82 0 27 60% of what they would eat given free access to food are

Downloaded from http://biomedgerontology.oxfordjournals.org/ at Rutgers University Libraries/Technical Services on April 10, 2015
ES1 13 13 62 0 54
found to live about 25%–40% longer than ad libitum-fed
ES2 15 11 64 0 36
ES3 12 10 50 0 20
controls. The improvement in longevity is accompanied by
LB1 13 9 78 0 33 a deceleration of age-dependent changes in many cell types
LB2 5 5 20 0 20 and organ systems, and retards the appearance of a wide
LB3 3 3 67 0 0 range of lethal and nonlethal diseases; this is consistent with
LS1 14 14 79 14 7 the idea that the intervention has retarded or decelerated a
LS2 15 13 54 8 15 fundamental aging process that helps to time the course of a
LS3 9 7 43 0 14
Sums: 181 156
wide range of changes in middle age and later life. Limita-
Mean (for stocks) 64 7 18 tion of dietary methionine to the minimal levels needed for
Proportion test§ p ⫽ .17 p ⬍ .001 p ⫽ .056 survival is also reported to give a 30%–45% increase in lon-
gevity in F344 rats (5,17); rats exposed to this diet are 43%
*The number of animals submitted for necropsy.
†The number of necropsies for which an unambiguous cause of death could
lighter than controls, and the extent to which this interven-
be assigned. tion mimics the metabolic and pathophysiological effects of
‡The proportion of cases for which death was attributed to neoplasia (including caloric restriction is still an open question.
cases of pituitary adenoma), as a percentage of all cases to which a cause of Natural and induced mutations that result in smaller body
death was assigned. size can also lead to dramatic increases in mouse longevity.
§The probability that the indicated cause of death is equally likely among all
Mice homozygous for the Ames dwarf mutation df/df, at the
15 stocks.
locus now known as Prop-1, live 50% to 70% longer than
normal sized, non-mutant controls (6), as do mice of the
Snell dwarf mutant, dw/dw, at the Pit-1 locus (18). Each of
pituitary adenoma. No other lethal cases of this lesion were these mutations induces the same essential change, i.e., a
seen among the other LB and EB lines, although pituitary loss, during embryogenesis, of signals needed to induce the
adenoma was lethal in 2 of 14 LS1 mice and in 1 of 13 LS2 cells of the anterior pituitary responsible for secretion of
mice. The proportions test confirmed ( p ⬍ .001) the im- growth hormone, thyroid-stimulating hormone, and prolac-
pression that this lethal lesion was distributed differentially tin. The decline in growth hormone and thyroid hormone se-
among the 15 stocks. The incidence of fatal lymphoma var- cretion prevents attainment of normal body size, and the
ied among the stocks from ⬍10% in five stocks to ⬎30% in mice are only about 35% of normal body weight at 8–12
three stocks; the proportions test provided marginal support weeks of age; both weight and linear dimensions are af-
( p ⫽ .056) for the hypothesis that the stocks differed among fected in parallel. Unless treated with growth hormone and/
themselves in the proportion of fatal cases of lymphoma. No or thyroid hormone, the mice remain small all of their long
attempt was made in these calculations to adjust for the po- lives, although they become quite obese in middle age.
tential confounds of competing hazards, although it is possi- Transgenic mice that overproduce urokinase-type plasmi-
ble that high early mortality attributable to one disease (e.g., nogen activator in brain consume 20% less food than litter-
pituitary adenoma in EB2) could lead to a corresponding mates, and exhibit a 20% decline in body size and a 20% in-
decline in the frequency of other possible causes of death. crease in longevity (7). It seems likely in this case that the
Stock longevity was not significantly related either to the genetic alteration is acting through an alteration in appetite
frequency of neoplasia (product moment r ⫽ –.13) or to the rather than a change in endocrine control of growth and de-
frequency of lymphoma (r ⫽ .12). velopment (as in the dw/dw and df/df mutants).
Artificial selection for suitability to a variety of tasks has
DISCUSSION produced a very wide variation in size among dog breeds,
These analyses suggest a significant correlated response of up to 36-fold when measured as body weight (8). Breeds
longevity to direct selection for differential rates of early also vary in mean longevity, from 6.9 to 10.8 years, and the
and late growth during the first 56 days of life in a set of 15 correlation between breed longevity and mean body weight
genetic strains of mice. This result is consistent with the is a remarkably high R2 ⫽ 56%. In some cases this variation
idea that polymorphic alleles with an impact on life span in size has been shown to be due to alterations in production
may have effects much earlier in the life history, and that of IGF-1, the principal mediator of growth hormone effect
their frequency in a population can be influenced by selec- (19,20). It is not known whether interbreed differences in
tive pressures acting on developmental, rather than adult or size are in all cases due to changes in the GH/IGF-1 axis,
senescent, phenotypes. Our data also add further support to nor how many loci contribute polymorphic alleles that af-
the hypothesis that longevity and body size are influenced at fect breed size and longevity, but the available data are con-
B460 MILLER ET AL.

sistent with the hypothesis that polymorphic loci affecting mice from day 0–10 represents an abnormality of pituitary
size have pleiotropic effects on dog life span. The evidence gland function that leads, later in life, to fatal adenoma forma-
on size/longevity correlations in humans is complicated by tion at a relatively early age (life span ⫽ 473 ⫾ 162 days).
a large number of potential confounding factors, including It has been shown elsewhere that selection for early rates
intergroup differences in health status, socioeconomic sta- of increased or decreased rate of growth in body weight re-
tus, childhood nutrition, and ancestry; in general, however, sulted in significant changes in cell number (i.e., hyperpla-
they are consistent with the associations noted in the animal sia, in the brain of these mice) (Atchley, Wei, and Crenshaw,
models. The effects are of substantial magnitude: in a study unpublished data). It thus seems likely that the selection
of 373 male veterans who died between 1984 and 1988, for process has acted on genes that participate in cell prolifera-
example, men whose height did not exceed 1.75 meters tion, some of which (such as N-myc) are known to affect

Downloaded from http://biomedgerontology.oxfordjournals.org/ at Rutgers University Libraries/Technical Services on April 10, 2015
were found (21) to live 4.95 years longer than taller mem- neural tissue preferentially. The high incidence of pituitary
bers of the study population, and those shorter than 1.7 adenoma in EB2 may be the result of the fixation during se-
meters were on average 7.5 years longer-lived than those lection of an aberrant form of this or some other oncogene.
taller than 1.83 meters. The absence of pituitary adenomas in the other five
In the current study we did not make any measurement of stocks of EB and LB mice suggest that there are likely to be
body composition, such as fat/lean ratio, but we consider it a number of other genetic combinations that can lead to
unlikely that the effects seen are due simply to interstock rapid weight gain either early or later in the first 56 days of
differences in obesity. For one thing, the correlation be- life but that act via pathways different from those that affect
tween stock body weight and stock longevity was as strong EB2 mice and thus do not lead to pituitary adenoma. The
(r ⫽ .59, p ⫽ .02) when the weight measure was derived data also include examples where stocks selected according
from 3-month-old mice as when the weight data were ob- to the same protocol differ significantly in life span (e.g.,
tained at peak body weight (r ⫽ .61, p ⫽ .02), or at 12 LS2 ⬎ LS1, LS3; and EB1, EB3 ⬎ EB2; and LB2, LB1 ⬎
months of age, i.e., at ages at which mice have become sub- LB3) or in weight (e.g., for peak weight: EB3 ⬎ EB1, EB2;
stantially more obese than at 3 months (r ⫽ .56, p ⫽ .03). It ES1, ES2 ⬎ ES3; LB3 ⬎ LB2, and LS1 ⬎ LS2, LS3).
is noteworthy in this context that the effect of caloric re- These mouse stocks may provide useful starting points
striction on longevity is equally apparent in mice of the ob/ for further investigations of the genetic control of growth
ob stock, whose lack of leptin expression makes them trajectory, body size, life span, and disease risk. Crosses
obese, compared to ad lib fed controls, even when on a ca- among the stocks would be expected to segregate alleles
lorically restricted diet (22). Similarly, the striking growth with effects on longevity, body size, and other life history
extension of the df/df and dw/dw dwarf mouse lines is ac- traits (such as litter size and maturation rate), and provide
companied by progressive obesity in midlife and at late insights into the number, location, and effect size of alleles
ages. Furthermore, the strong association between body that influence one or more of these traits. Mice of the LS2
weight and longevity among dog breeds is not attributable stock, in particular, may be a source of allele combinations
to parallel interbreed differences in obesity. In 12 of our set that convey exceptional longevity compared to those com-
of 15 mouse stocks the interstock differences in adult and mon among typical laboratory inbreds. We have in previous
midlife body weight are very likely to reflect selection- studies used mice of the UM-HET3 stock, bred as the prog-
driven alterations in allele frequencies that influence either eny of (BALB/c ⫻ C57BL/6)F1 dams and (C3H ⫻ DBA/
the rate of weight gain from day 0–10 (ES and EB stocks) or 2)F1 sires, to map quantitative trait loci associated with dif-
weight gain from day 28–56 (LS and LB stocks), although it ferential longevity (23). Females of the UM-HET3 stock,
is hard to rule out effects of genetic drift that have led to housed in the same vivarium over the same time interval,
substantial (and significant) differences in weight and life were significantly shorter lived than the LS2 females pre-
span in one of the three unselected control lines (i.e., C3). sented in this report (UM-HET3 life span was 806 ⫾ 167
It is unlikely that these differences reflect selection for obe- days, N ⫽ 148, p ⫽ .003 compared to LS2 females by two-
sity in the first 10 or 56 days of life, although formal studies tailed t test). Further work with the LS2 mice will be needed
of body composition would be needed to address this possi- to see if their superior longevity can be replicated in other
bility. environments, affects males as well as females, and associ-
Each of the four selection protocols (EB, LB, LS, ES) ates in backcross generations with small body size and/or
was applied to generate three independent replicate stocks altered patterns of fertility.
that share no common ancestor subsequent to the applica-
tion of the selection pressure. The selection procedures were Acknowledgments
carried out on closed populations (i.e., there was no inter- This research was supported by the Nathan Shock Center for the Biology
breeding between replicate lines within each selection pro- of Aging, National Institutes of Health Grants AG13283 and GM-45344 to
tocol). It is interesting to note that the replicate mouse William Atchley. We thank Luann Linsalata and Gretchen Buehner for
technical assistance, and Dr. Maria Moalli for veterinary supervision.
stocks thus produced can in some cases vary considerably
from one another in pathology, peak weight, and longevity. Address correspondence to Dr. Richard A. Miller, Room 5316 CCGCB,
Box 0940, University of Michigan, 1500 East Medical Center Drive, Ann
The most dramatic example comes from the incidence rates Arbor, MI 48109-0940. E-mail: [email protected]
of pituitary adenoma, which led to the death of 84% of the
mice in stock EB2 but which were not noted in any of the References
necropsies of EB1, EB3, or any of the LB stocks. It is plau- 1. Rose MR. Evolutionary Biology of Aging. New York: Oxford Univer-
sible, though unproven, that the rapid growth rate of EB2 sity Press; 1991.
 BODY SIZE AND LONGEVITY IN MICE B461

2. Westendorp RG, Kirkwood TB. Human longevity at the cost of repro- rate on reproductive onset in female mice. Genet Res. 1999;74:55–64.
ductive success. Nature. 1998;396:743–746. 15. Chrisp CE, Turke P, Luciano A, Swalwell S, Peterson J, Miller RA.
3. Austad SN. Why We Age: What Science is Discovering About the Lifespan and pathology in genetically heterogeneous (four-way cross)
Body’s Journey Through Life. New York: John Wiley & Sons; 1997. mice: a new model for aging research. Vet Pathol. 1996;33:735–743.
4. Weindruch, R, Walford RL. The Retardation of Aging and Disease by 16. Sokal RR, Rohlf FJ. Biometry: The Principles and Practice of Statis-
Dietary Restriction. Springfield, IL: Charles C Thomas; 1988. tics in Biological Research. 3rd ed. New York: Freeman; 1995.
5. Orentreich N, Matias JR, DeFelice A, Zimmerman JA. Low meth- 17. Richie JP Jr, Leutzinger Y, Parthasarathy S, Malloy V, Orentreich N,
ionine ingestion by rats extends life span. J Nutr. 1993;123:269–274. Zimmerman JA. Methionine restriction increases blood glutathione
6. Brown-Borg HM, Borg KE, Meliska CJ, Bartke A. Dwarf mice and and longevity in F344 rats. FASEB J. 1994;8:1302–1307.
the ageing process. Nature. 1996;384:33. 18. Miller RA. Genes for ageing? Trends Genet. 1999;15:175–176.
7. Miskin R., Masos T. Transgenic mice overexpressing urokinase-type 19. Eigenmann JE, Patterson DF, Froesch ER. Body size parallels insulin-
plasminogen activator in the brain exhibit reduced food consumption, like growth factor I levels but not growth hormone secretory capacity.

Downloaded from http://biomedgerontology.oxfordjournals.org/ at Rutgers University Libraries/Technical Services on April 10, 2015
body weight and size, and increased longevity. J Gerontol Biol Sci. Acta Endocrinol. 1984;106:448–453.
1997;52A:B118–B124. 20. Eigenmann JE, Amador A, Patterson DF. Insulin-like growth factor I
8. Li Y, Deeb B, Pendergrass W, Wolf N. Cellular proliferative capacity levels in proportionate dogs, chondrodystrophic dogs and in giant
and life span in small and large dogs. J Gerontol Biol Sci. 1996;51A: dogs. Acta Endocrinol. 1988;118:105–108.
B403–B408. 21. Samaras TT, Storms LH. Impact of height and weight on life span.
9. Hillesheim E, Stearns SC. Correlated responses in life-history traits to Bull World Health Organization. 1992;70:259–267.
artificial selection for body weight in Drosophila melanogaster. Evo- 22. Harrison DE, Archer JR, Astle CM. Effects of food restriction on ag-
lution. 1992;46:745–752. ing: separation of food intake and adiposity. Proc Natl Acad Sci USA.
10. Atchley WR, Xu S, Cowley DE. Altering developmental trajectories in 1984;81:1835–1838.
mice by restricted index selection. Genetics. 1997;146:629–640. 23. Miller RA, Chrisp C, Jackson AU, Burke DT. Marker loci associated
11. Enesco M, Leblond EP. Increase in cell number as a factor in the with lifespan in genetically heterogeneous mice. J Gerontol Med Sci.
growth of the organs and tissues of the young male rat. J Embryol Exp 1998;53A:M257–M263.
Morph. 1962;10:530–562.
12. Goss RJ. Hypertrophy and hyperplasia. Science. 1966;153:1615–1620.
13. Falconer DS, Gauld IK, Roberts RC. Cell numbers and cell sizes in or-
gans of mice selected for large and small body size. Genet Res. 1978; Received August 6, 1999
31:287–301. Accepted March 10, 2000
14. Ernst CA, Crenshaw PD, Atchley WR. Effect of selection for development Decision Editor: Jay Roberts, PhD