Darwin was right: inbreeding depression on male fertility in the Darwin family
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
bs_bs_banner
Biological Journal of the Linnean Society, 2015, 114, 474–483. With 2 figures
Darwin was right: inbreeding depression on male
fertility in the Darwin family
GONZALO ÁLVAREZ1, FRANCISCO C. CEBALLOS*1 and TIM M. BERRA FLS2,3
1
Department of Genetics, Faculty of Biology, University of Santiago de Compostela, 15782 Santiago
de Compostela, Spain
2
Department of Evolution, Ecology and Organismal Biology, The Ohio State University, 1760
University Dr., Mansfield, OH 44906, USA
3
Research Institute for the Environment and Livelihoods, Charles Darwin University, Darwin, NT,
Australia
Received 25 July 2014; revised 3 September 2014; accepted for publication 4 September 2014
Charles Darwin, who was married to his first cousin Emma Wedgwood, was the first experimentalist to
demonstrate the adverse effects of inbreeding. He documented the deleterious consequences of self-fertilization on
progeny in numerous plant species, and this research led him to suspect that the health problems of his 10
children, who were very often ill, might have been a consequence of his marriage to his first cousin. Because
Darwin’s concerns regarding the consequences of cousin marriage on his children even nowadays are considered
controversial, we analyzed the potential effects of inbreeding on fertility in 30 marriages of the Darwin–Wedgwood
dynasty, including the marriages of Darwin’s children, which correspond to the offspring of four cousin marriages
and three marriages between unrelated individuals. Analysis of the number of children per woman through
zero-inflated regression models showed a significantly adverse effect of the husband inbreeding coefficient on family
size. Furthermore, a statistically significant adverse effect of the husband inbreeding coefficient on reproductive
period duration was also detected. To our knowledge, this is the first time that inbreeding depression on male
fertility has been detected in humans. Because Darwin’s sons had fewer children in comparison to non-inbred men
of the dynasty, our findings give empirical support to Darwin’s concerns on the consequences of consanguineous
marriage in his own progeny. © 2014 The Linnean Society of London, Biological Journal of the Linnean Society,
2015, 114, 474–483.
ADDITIONAL KEYWORDS: human fertility – zero-inflated regression models.
INTRODUCTION mental research programme on the harmful effects of
inbreeding was performed by Charles Darwin, who
Scientific and family concerns relative to inbreeding
carried out carefully controlled experiments that
converge in Charles Darwin’s biography. Inbreeding
involved self-fertilization and outcrossing between
is usually defined as the mating between relatives
unrelated individuals in 57 plant species (Darwin,
and leads to increased homozygosity in the progeny of
1868, 1876). In these experiments, Darwin docu-
such a mating. In humans, genome-wide scans show
mented the phenomenon of inbreeding depression
that inbred individuals are characterized by numer-
because the offspring of self-fertilized plants were on
ous long chromosomal segments of marker
average shorter, flowered later, weighed less, and
homozygosity (termed ROHs, runs of homozygosity),
produced fewer seeds than the progeny of cross-
which appear to be randomly distributed along their
fertilized plants. Darwin’s laborious study on inbreed-
chromosomes (Gibson, Morton & Collins, 2006; Woods
ing had its origin in his interest on plant reproductive
et al., 2006; McQuillan et al., 2008). The first experi-
systems because his research was performed to
explain why numerous plant species have systems
that prevent self-fertilization and why reproduction
*Corresponding author. E-mail: francisco.ceballos@usc.es by outcrossing is prevalent in nature (Pannell, 2009).
474 © 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483INBREEDING DEPRESSION IN THE DARWIN FAMILY 475
However, it is very likely that Darwin also had a Nowadays, Charles Darwin’s concerns on the
personal interest in the matter. Charles Darwin was harmful effects of first-cousin marriage in his progeny
married to his first cousin Emma Wedgwood and they are often considered unjustified or, at least, exagger-
had 10 children (Freeman, 1982; Browne, 2002; Berra, ated because they were based on the extrapolation
2013). Darwin, who suffered illness for most of his from ill-effects of self-fertilization (inbreeding coeffi-
adult life with many differing symptoms (Colp, 2008; cient, F = 0.50) in plants to the outcomes of first-
Hayman, 2013), was worried about the health of his cousin marriage (F = 0.0625) in humans, as well as on
children, who were very often ill, and three of them prejudices against consanguineous marriage preva-
died before adulthood: Anne Elizabeth (1841–51), lent in that time (Jones, 2008; Bittles, 2009). Never-
Mary Eleanor (1842), and Charles Waring (1856–58). theless, the possibility of an adverse effect of
Darwin’s own ill health not only led him to fear that his inbreeding on fertility in the offspring of a number of
children could have inherited his medical problems, cousin marriages of the Darwin–Wedgwood dynasty
but also he suspected that his marriage to his first has been repeatedly pointed out (Moore, 2005;
cousin might have caused some of his children’s health Golubovsky, 2008). Three of Charles Darwin’s six
problems (Browne, 2002; Moore, 2005; Jones, 2008; children with long-term marriage history (William,
Bittles, 2009; Kuper, 2009; Berra, 2013). The interest Henrietta, and Leonard) had no progeny and their
of Darwin on the consequences of human inbreeding unexplained infertility might have been the result of
led him to ask his friend John Lubbock, member of increased homozygosity for recessive autosomal
Parliament, to make a request to Parliament for the meiotic mutations as a result of cousin marriage
inclusion of a question on consanguineous marriage in (Golubovsky, 2008). In the same sense, it has been
the 1871 Census of Great Britain and Ireland (Browne, also noted that a number of individuals of the
2002; Bittles, 2009; Berra, Álvarez & Ceballos, 2010a). Darwin–Wedgwood dynasty, including the offspring of
Charles Darwin’s son George was also involved in Emma Wedgwood and her brothers Josiah III, Henry,
the matter. He performed a study on cousin marriage and Hensleigh, who were also married to cousins,
in England, concluding that the adverse effects of presented low fertility (Moore, 2005). However, these
consanguineous marriage could be not so strong as observations do not constitute convincing evidence of
assumed in that time, particularly in the best families: an adverse effect of inbreeding on fertility in the
‘I may mention that Dr. Arthur Mitchell, of Edinburgh, Darwins’ children because the relationship between
conducted an extensive inquiry, and came to the con- inbreeding and fertility among marriages has not
clusion that, under favourable conditions of life, the been investigated in the Darwin–Wedgwood dynasty.
apparent ill-effects were frequent almost nil, whilst if Furthermore, it is necessary to take into account that
the children were ill fed, badly housed and clothed, the the present knowledge of the impact of inbreeding on
evil might become very marked. This is in striking fertility is very limited in humans. By contrast to the
accordance with some unpublished experiments of my extensive evidence for inbreeding depression on pre-
father, Mr. Charles Darwin, on the in-and-in breeding reproductive survival (Bittles & Black, 2010; Álvarez,
of plants; for he has found that in-bred plants, when Quinteiro & Ceballos, 2011; Bittles, 2012; Ceballos &
allowed enough space and good soil, frequently show Álvarez, 2013), the effects of increased homozygosity
little or no deterioration, whilst when placed in com- on human fertility caused by inbreeding are little
petition with another plant, they frequently perish or known and only a few studies have reported conclu-
are much stunted’ (Darwin, 1875). Darwin was very sive evidence (Ober, Hyslop & Hauck, 1999; Robert
influenced by George’s research in such a way that he et al., 2009; Postma, Martini & Martini, 2010). The
revised his opinion on the effects of consanguineous high incidence of cousin marriages in the Charles
marriage in his late years. In the last edition of The Darwin family gives the opportunity of using such
variation of animals and plants under domestication, marriages as a useful framework for investigating the
published in 1875, Darwin claimed: ‘Whether consan- effects of inbreeding on human fertility, which, in
guineous marriages, such as are permitted in civilized turn, could shed light on Darwin’s concerns regarding
nations, and which would not be considered as close the consequences of consanguineous marriage in his
interbreeding in the case of our domesticated animals, own progeny.
cause any injury will never be known with certainty
until a census is taken with this object in view. My son,
George Darwin, has done what is possible at present by
MATERIAL AND METHODS
a statistical investigation, and he has come to the
conclusion, from his own researches and those of Dr. GENEALOGICAL AND DEMOGRAPHIC DATA
Mitchell, that the evidence as to any evil thus caused Genealogical information obtained from The Exciting
is conflicting, but on the whole points to the evil being Wedgwoods Home Page (http://www.familyhistorian
very small’. .info/exciting/wedgwood/index.html), Kindred Britain
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483476 G. ÁLVAREZ ET AL.
Figure 1. Partial pedigree of the 26 individuals of the Darwin–Wedgwood dynasty considered for the fertility
analysis.
(http://www.kindred.stanford.edu), and other sources depth of at least five generations. Inbreeding coeffi-
(Freeman, 1984; Berra, Álvarez & Shannon, 2010b) cients for individuals belonging to other different
was used to extend the pedigree of the Darwin– families who were married to Darwin–Wedgwood
Wedgwood dynasty constructed by Berra et al. individuals were based on at least three generations,
(2010a) to include all the individuals considered in and therefore they were individuals for whom all
the present study (Fig. 1). Pedigree analysis was used their great-grandparents were known.
to calculate the individual inbreeding coefficient for Demographic data consisted of the total number of
husband (Fh) and wife (Fw) and the kinship coefficient children produced per woman, the age of marriage for
(θ) of the couple by means of FSPEED software both husband and wife, the duration of marriage, the
(Tenset Technologies; http://www.tenset.co.uk/fspeed). protogenesic interval (time interval between marriage
The inbreeding coefficient (F) is the probability that and birth of the first child), the intergenesic interval
an individual receives at a given autosomal locus two (mean time between two successive birth events), and
alleles that are identical by descent or, equivalently, the reproductive span (time interval between the first
the proportion of the individual’s autosomal genome and the last child). To obtain a measure of the repro-
expected to be homozygous by descent (autozygous) ductive period duration for all women irrespective of
(Cavalli-Sforza & Bodmer, 1971; Falconer & Mackay, the number of children, an index denoted effective
1996; Hedrick, 2011). The proportion of the individu- reproductive span was defined as the reproductive
al’s genome that is identical by descent (f) is expected span + 1. In this way, a couple with one child has an
to be the inbreeding coefficient (F) with variance effective reproductive span of 1 year, and an effective
Var (f) ≈ 2F(1 – F)/ρG, where G is the total length of reproductive span of zero was assigned to couples
the autosomal genome expressed in Morgans and with no children. The demographic information
ρ = nm + nf, where nm and nf denote, respectively, was obtained from the genealogical sources above
the number of meiosis from the individual to the mentioned.
ancestral pair in the paternal and maternal lines
(Carothers et al., 2006). The coefficient of kinship of a
couple is the probability that two alleles at the same STATISTICAL ANALYSIS
locus drawn at random, one from each spouse, are The effects of explanatory variables such as Fh, Fw, θ,
identical by descent, and therefore the inbreeding age at marriage for husband and wife, and duration of
coefficient of an individual is equal to kinship coeffi- marriage on the total number of children per woman
cient of his or her parents. The amount of pedigree were investigated through different regression models
information available for the individuals from the for count data: generalized linear models (GLMs) and
Darwin–Wedgwood dynasty allowed us to compute zero-inflated models (Zeileis, Kleibe & Jackman,
inbreeding coefficients on the basis of a pedigree 2008; Zuur et al., 2009). Zero-inflated Poisson (ZIP)
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483INBREEDING DEPRESSION IN THE DARWIN FAMILY 477 and zero-inflated negative binomial (ZINB) regression RESULTS models, as well as their classical counterparts, Poisson GLM and negative binomial GLM, were used The inbreeding effects on fertility in the Darwin– because the distribution of number of children from Wedgwood dynasty were investigated in the offspring Darwin–Wedgwood women was characterized by of a number of marriages that presented remarkable excess zeros and overdispersion (sample variance differences in their degree of consanguinity (Fig. 1). larger than the mean). The relative goodness of fit of Four cousin marriages were contracted by Charles the different regression models to data was evaluated Darwin’s wife, Emma Wedgwood, and her brothers by the Akaike information criterion (AIC), which was Josiah III, Hensleigh and Henry. Hensleigh used as a model selection criterion to decide on the Wedgwood was married to his first cousin, Frances optimal model. The negative binomial distribution is Makintosh, giving place to offspring with an inbreed- commonly used for overdispersed count data in many ing coefficient of 0.0625. The marriage of Charles and areas of biological research (White & Bennetts, 1996; Emma and the marriage of Josiah Wedgwood III and Lloyd-Smith, 2007) and is usually expressed in terms Caroline Darwin, Charles Darwin’s sister, were also of the mean (m) and a dispersion parameter k, and first-cousin unions but their progeny had an inbreed- its variance is m + m2/k. The smaller k, the larger ing coefficient of 0.0630 given that their grandpar- the overdispersion. If k→∞, the negative binomial ents, Josiah Wedgwood I and Sarah Wedgwood, were converges to the Poisson distribution with variance third cousins. Henry Wedgwood and his wife Jessie equal to m. Negative binomial GLM can cope with Wedgwood were double first cousins because their overdispersion as a result of extra variation in fathers, Josiah Wedgwood II and John Wedgwood, the nonzero part of the data, whereas zero-inflated were brothers and their mothers, Elizabeth Allen and regression models are capable of dealing with Louisa Allen, were sisters, in such a way that their overdispersion as a result of an excessive numbers of progeny had an inbreeding coefficient of 0.1255, zeros (Zeileis et al., 2008; Zuur et al., 2009). Zero taking into account that their paternal grandparents inflation models are two-component mixture models were third cousins. On the other hand, two other that combine a point mass at zero with a count Emma’s siblings, Charlotte and Francis, as well as distribution such as Poisson or negative binomial. In their cousin, Robert Wedgwood, contracted marriages these models, the zeros are modelled as coming from to unrelated individuals (Charles Langton, Frances two different processes: the true zeros correspond to Mosley, and Mary Hasley, respectively) and, there- the count process and are modelled by a Poisson (ZIP) fore, in these three cases, the progeny had an inbreed- or negative binomial (ZINB) GLM, whereas the false ing coefficient equal to zero. The offspring of these zeros are modelled by a binomial GLM. It is reason- seven couples involved a total number of 26 married able to assume that zero inflation models are very individuals (13 men and 13 women) who had a appropriate for our particular situation because common genetic background, the Wedgwood back- a significant number of women (14/30; 46.7%) had ground, and presented remarkable differences in their zero children, in such a way that some of them inbreeding level (ranging from F = 0 to F = 0.1255). could be biologically sterile and therefore they could These 26 individuals contracted a total number of 30 correspond to false zeros. On the other hand, the marriages (Table 1), taking into account that one man likelihood-ratio test was used to choose between married three times (Darwin’s son Francis), three Poisson and negative binomial models and between men married twice [Darwin’s son Leonard, Godfrey ZIP and ZINB because Poisson and negative binomial Wedgwood (Francis Wedgwood’s son) and Rowland are nested models and the same is true for ZIP and Wedgwood (Henry Wedgwood’s son)], and that ZINB. Estimates of the dispersion parameter k of the Godfrey Wedgwood’s second wife was Hope Wedgwood negative binomial distribution were obtained by the (Hensleigh Wedgwood’s daughter). These 30 mar- maximum likelihood method and the likelihood-ratio riages were considered for the fertility analysis and test above mentioned was used for testing the included women born in the period from 1834 to 1875 statistical significance of a k estimate. The effective and men born in the period 1832–1854. reproductive span was considered as a count response The inbreeding coefficient for husband and wife (Fh variable and therefore it was analyzed by means of and Fw) and the kinship of couple (θ), as well as the the regression models for the count data mentioned demographic variables number of children per woman above. Protogenesic and intergenesic intervals were (family size), age at marriage for husband and response variables investigated through Gaussian wife (AMh and AMw), duration of marriage (DM), linear regression by ordinary least squares. All protogenesic and intergenesic intervals, and repro- the statistical analyses were conducted using the ductive span from the 30 marriages of the Darwin– statistical software R (R Development Core Team, Wedgwood dynasty are given in Table 1. Mean ± SE 2011). values for Fh, Fw, and θ were 0.0437 ± 0.0092, © 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483
Table 1. Inbreeding coefficient for husband and wife (Fh and Fw), kinship coefficient of couple (θ), number of children, age at marriage for husband and wife (AMh,
478
AMw), duration of marriage (DM), protogenesic interval, intergenesic interval, and reproductive span from 30 marriages of the Darwin–Wedgwood dynasty
Number of Protogenesic Intergenesic Reproductive
Husband Wife Fh Fw θ children AMh AMw DM interval interval span
Charles Darwin & Emma Wedgwood’s children
William Darwin Sara Sedgwick 0.0630 0.0000 0.0000 0 38 38 25 – – –
Richard Litchfield Henrietta Darwin 0.0000 0.0630 0.0000 0 39 27 32 – – –
George Darwin Maud du Puy 0.0630 0.0000 0.0000 5 39 23 28 12 42.00 14
Francis Darwin Amy Ruck 0.0630 – – 1 26 26 2 28 – 0
Francis Darwin Ellen Wordsworth 0.0630 – – 1 35 27 20 31 – 0
G. ÁLVAREZ ET AL.
Francis Darwin Florence Fisher 0.0630 0.000 0.0000 0 65 48 7 – – –
Leonard Darwin Elisabeth Fraser 0.0630 0.0000 0.0392 0 32 32 16 – – –
Leonard Darwin Charlotte Massingberd 0.0630 0.0000 0.0000 0 50 36 40 – – –
Horace Darwin Emma Farrer 0.0630 0.0000 0.0000 3 29 25 48 22 23.00 4
Josiah Wedgwood III & Caroline Darwin’s children
Arthur Williams Margaret Wedgwood 0.0000 0.0630 0.0000 3 33 24 5 19 18.00 3
Matthew Harrison Lucy Wedgwood – 0.0630 – 3 28 27 45 21 15.00 3
Charles Langton & Charlotte Wedgwood′s children
Edmund Langton Caroline Massingberd 0.000 0.000 0.0625 4 26 19 8 14 20.00 8
Francis Wedgwood & Frances Mosley’s children
Godfrey Wedgwood Mary Hawkshaw 0.0000 – – 1 29 26 1 12 – 0
Godfrey Wedgwood Hope Wedgwood* 0.0000 0.0625 0.0782 1 43 32 29 48 – 0
John Hawkshaw Cecily Wedgwood – 0.0000 – 4 24 28 52 12 26.33 9
Clement Wedgwood Emilie Rendel 0.0000 0.0000 0.0000 6 26 26 23 12 24.80 10
Lawrence Wedgwood Emma Houseman 0.0000 – – 6 27 30 42 27 19.80 9
Johannes Franke Constance Wedgwood – 0.0000 – 0 33 34 23 – – –
Arthur Parson Fanny Wedgwood – 0.0000 – 0 35 28 27 – – –
Henry Wedgwood & Jessie Wedgwood’s children
William Kempson Louisa Wedgwood – 0.1255 – 4 30 30 17 35 30.33 7
John Wedgwood Helen Tyler 0.1255 0.0000 0.0000 2 26 21 16 18 28.00 3
Ralph Carr Anne Wedgwood – 0.1255 – 0 37 29 7 – – –
Rowland Wedgwood Sophia Rudd 0.1255 – – 0 36 26 16 – – –
Rowland Wedgwood Agnes Harley 0.1255 – – 0 60 32 14 – – –
Hensleigh Wedgwood & Frances Makintosh’s children
Ernest Wedgwood Mary Bell 0.0625 – – 1 50 22 11 74 – 0
Thomas Farrer Katherine Wedgwood – 0.0625 – 0 34 33 26 – – –
Alfred Wedgwood Margaret Ingall 0.0625 0.0000 0.0000 3 31 19 19 36 96.00 16
Robert Wedgwood & Mary Hasley’s children
Reginald Hoskins Eleanor Wedgwood 0.0000 0.0000 0.0000 0 30 33 3 – – –
Clement Allen Edith Wedgwood 0.0000 0.0000 0.0313 0 33 23 13 – – –
Wilfred Allen Anna Wedgwood 0.0000 0.0000 0.0313 0 34 28 39 – – –
Mean 0.0437 0.0246 0.0152 1.60 35.3 28.4 21.8 26.3 31.2 5.38
SE 0.0092 0.0086 0.0064 0.36 1.8 1.1 2.6 4.1 6.8 1.30
*Hensleigh Wedgwood’s daughter.
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483
Inbreeding/coancestry coefficient was not computed for those individuals/marriages with incomplete genealogical information, which is indicated by a hyphen.INBREEDING DEPRESSION IN THE DARWIN FAMILY 479
0.0246 ± 0.0086 and 0.0152 ± 0.0064, respectively. Table 2. Regression analysis of number of children per
The distribution of the number of children per woman woman as a function of husband and wife inbreeding
presented a relatively low mean ± SE, 1.60 ± 0.36, coefficient (Fh and Fw), kinship (θ), age at marriage for
mainly as a result of the high proportion of women husband and wife (AMh and AMw) and duration of mar-
who had no children (14/30, 46.7%). In addition, the riage (DM) through a zero-inflated Poisson (ZIP) model
distribution was overdispersed because the observed
ratio of variance to mean was 2.44 (3.90/1.60), a value Coefficient SE P
that is not very different from the ratio found in many
Regression analysis for 16 marriages (complete model)
human populations where variances in the number of
Fh −14.806 ±8.979 0.0496
progeny are from 1.5- to three-fold as great as their
Fw −1.388 ±13.335 0.459
means (Cavalli-Sforza & Bodmer, 1971). Indeed, the
θ −12.834 ±9.724 0.093
observed distribution significantly departs from the AMh −0.002 ±0.053 0.489
Poisson distribution with the same mean (χ2 = 24.39, AMw −0.172 ±0.075 0.011
d.f. = 3, P < 0.001), and it was mainly the result of DM 0.027 ±0.023 0.120
excessive number of zero values because the number Regression analysis for 23 marriages (Fw and θ removed)
of women who had no children was 14, whereas the Fh −12.546 ±5.986 0.018
expected value according to Poisson was 6.1. The AMh −0.031 ±0.036 0.195
mean ± SE duration of the Darwin–Wedgwood mar- AMw −0.146 ±0.051 0.002
riages was 21.8 ± 2.6 years, and mean ± SE values for DM 0.041 ±0.014 0.002
age at marriage for husband and wife were 35.3 ± 1.8
years and 28.4 ± 1.1 years, respectively. The repro-
ductive span of Darwin–Wedgwood women was
5.38 ± 1.30 years, an extremely short reproductive GLM, respectively). Indeed, the dispersion parameter
period in comparison to values of contemporary popu- k of the negative binomial distribution was not sta-
lations such as the Saguenay-Lac-Saint-Jean popula- tistically significant by the likelihood-ratio test in
tion in Canada, where recent research focusing on both ZINB and negative binomial GLM. Conse-
182 women born in 1879 showed a mean reproductive quently, the corresponding results for the ZIP model
period duration of 15.87 years (Robert et al., 2009). are given in Table 2. A statistically significant adverse
Mean ± SD values for protogenesic and intergenesic effect of the inbreeding coefficient of husband (Fh) on
intervals (both in months) were 26.3 ± 4.1 and the number of children was found (P = 0.0496). In
31.2 ± 6.8, respectively. addition, a significantly negative effect of AMw on
The effects of inbreeding (Fh and Fw) and kinship (θ) family size was also detected. Those women who got
on the number of children per woman were investi- married at an earlier age had a higher number of
gated through ZIP and ZINB regression models, as progeny. The remaining explanatory variables did not
well as their classical counterparts, Poisson and nega- have a statistical significant effect on the number of
tive binomial GLMs (Zeileis et al., 2008; Zuur et al., children. To maximize statistical power, we subse-
2009). These regression models for count data were quently tested for an effect Fh on the number of
used because the distribution of number of children in children per woman by removing both Fw and θ from
the Darwin–Wedgwood women was characterized by the analysis, in such a way that we had a sample
a high proportion of women who had no children and size of 23 couples in the new regression analysis
overdispersion, as noted above. We first tested simul- (see Supporting information, Table S1). According to
taneously for an effect of Fh, Fw, θ, AMh, AMw, and DM the AIC criterion, the best model was ZIP and the
on the number of children per woman through ZIP, next best was Poisson GLM (AICs were 72.511 and
ZINB, Poisson GLM, and negative binomial GLM (see 73.358, respectively), as in the previous analysis.
Supporting information, Table S1). Note that the Accordingly, Table 2 (bottom) shows the results for
sample size (N = 16 couples) is greatly reduced for the ZIP model. A significantly negative effect of Fh on
this analysis because only couples for which we have the number of children was detected (P = 0.018). It is
sufficient pedigree information for both husband and therefore confirmed that inbred husbands had signifi-
wife (and their kinship) could be included. According cantly fewer children (Fig. 2). A significantly negative
to the AIC, the model that better fit to data was ZIP effect of AMw on family size was also detected, as in
and the second better was Poisson GLM, although the the previous analysis and, in addition, a significantly
difference between the two models was very small positive effect of DM on the number of children
(AICs were 58.072 and 58.530, respectively). The was also found. Because, in our data set, four men
regression models based on the negative binomial married more than once (Francis and Leonard
distribution presented a poor fit to data (AICs were Darwin and Godfrey and Rowland Wedgwood), a
60.073 and 60.295 for ZINB and negative binomial pseudoreplication problem could be present in our
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483480 G. ÁLVAREZ ET AL.
Table 3. Regression analysis of effective reproductive
span as a function of husband and wife inbreeding coeffi-
cient (Fh and Fw), kinship (θ), age at marriage for husband
and wife (AMh and AMw) and duration of marriage (DM)
through a zero-inflated Poisson (ZIP) model
Coefficient SE P
Fh −16.732 ±5.748 0.002
Fw −15.576 ±10.967 0.078
θ −21.713 ±8.143 0.004
AMh 0.019 ±0.033 0.281
AMw −0.243 ±0.065 0.009
DM 0.016 ±0.019 0.197
binomial GLM. ZIP was the regression model that
better fit to data in terms of the AIC values and the
corresponding results are given in Table 3 (the results
for the four regression models are provided in
the Supporting information, Table S3). Significantly
negative effects of both Fh and AMw on the effective
reproductive span were found (P = 0.002 and
P = 0.009, respectively). The higher inbreeding coeffi-
cient for husband, the lower reproductive period dura-
tion (Fig. 2) and, in the same way, the increase in
AMw led to lower effective reproductive span. In addi-
tion, a significantly negative effect of kinship of couple
(θ) on the effective reproductive span was also found.
Figure 2. Mean number of children and effective repro- The higher kinship of the couple, the lower effective
ductive span for different values of husband inbreeding reproductive span. Finally, it is interesting to note
coefficient (Fh) from 23 Darwin–Wedgwood marriages. The that the effects of female inbreeding (Fw) were not
bars corresponding to Fh = 0.0630 represent fertility detected on either the number of children per woman
values for Charles Darwin’s son. or effective reproductive span. It could be that female
inbreeding had no effect on fertility in the Darwin–
Wedgwood marriages, although it could be simply a
regression analyses. To circumvent this statistical consequence of the low inbreeding of females com-
problem, we performed a new regression analysis by pared to males in these marriages (the mean inbreed-
using the mean number of children per woman for ing for men was nearly twice as high as for women:
each one of those four men. The results obtained were 0.0437 versus 0.0246).
very similar to those from the previous analysis
(results not shown). Thus, ZIP gave a better fit
than Poisson GLM in terms of AIC values (68.81
DISCUSSION
versus 69.90, respectively) and a statistical signifi-
cant adverse effect of Fh on number of children was An adverse effect of male inbreeding on both number
detected (Fh = −12.257, P = 0.035). of children per woman and duration of reproductive
The effects of inbreeding on protogenesic and period was detected through zero-inflated regression
intergenesic intervals were investigated through models from 30 Darwin–Wedgwood marriages. Inbred
regression analyses by ordinary least squares. The men had significantly fewer children and shorter
analyses did not show statistical significant effects reproductive span (Fh = −12.546, P = 0.018 for
for any explanatory variable (Fh, Fw, θ, AMh, AMw, and number of children and Fh = −16.732, P = 0.002 for
DM) on either protogenesic or intergenesic interval effective reproductive span). To our knowledge, this is
(see Supporting information, Table S2). On the other the first time that inbreeding depression on male
hand, inbreeding effects on effective reproductive fertility has been detected in humans. Although a
span were investigated through regression models for number of studies have investigated the effect of
count data: ZIP, ZINB, Poisson GLM and negative kinship or consanguinity between spouses on the total
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483INBREEDING DEPRESSION IN THE DARWIN FAMILY 481
number of offspring (Bittles et al., 2002; Helgason number of long homozygous segments (exceeding 3
et al., 2008; Bittles, 2012), the impact of male and/or cM) randomly distributed along their chromosomes
female inbreeding on human fertility has not been (Woods et al., 2006). The number of homozygous
examined intensively and only a few studies have segments ranged from seven to 32 segments per
produced conclusive evidence. Thus, inbred women, individual (mean of 20 homozygous segments), and
but not inbred men, showed significantly fewer chil- the proportion of the individual genome that was
dren in both the Hutterites from South Dakota and a homozygous varied from 5% to 20%, with a mean
small and isolated Swiss village (Ober et al., 1999; value of 11%. Therefore, the variation in family size
Postma et al., 2010). On the other hand, although among Darwin’s sons is not unexpected from a
male inbreeding had no effect on total number of genomic perspective of inbreeding. On the other hand,
offspring, high levels of male inbreeding were associ- a significantly positive association between childhood
ated with a reduction of the productivity of parents mortality and inbreeding coefficient among the off-
during the second half of their reproductive period spring of 25 Darwin–Wedgwood marriages that
compared to the first half, in a cohort of Canadian included Darwin’s children has been reported (Berra
women born in late 19th Century (Robert et al., 2009). et al., 2010a). This finding suggests that the high
It was found that inter-birth intervals increase with child mortality experienced by Darwin progeny (three
parental age and this increase was significantly of his 10 children died at age 10 years or younger)
stronger for most inbred males. An increase in inter- with respect to the mortality of non-inbred progeny
birth interval with age stronger for most inbred males (9.34 ± 3.23) from other Darwin–Wedgwood families
was observed in the Darwin–Wedgwood marriages might be a result of increased homozygosity of del-
(results not presented), although this effect was not eterious alleles due to cousin marriage (Berra et al.,
statistically significant because the number of women 2010a). The reasoning that inbreeding was involved
who had three or more children was very small in our in childhood mortality in the Darwin progeny is con-
sample. The results reported in the present study for sistent with the cause of death for two of Darwin’s
the Darwin–Wedgwood dynasty are not unexpected children. Anne Elizabeth most likely died of child
from a wide perspective because inbreeding depres- tuberculosis (Keynes, 2001; Fenner, Egger &
sion on male fertility has been found in a number Gagneux, 2009) and Charles Waring died of scarlet
of animal species including mammals (Roldan et al., fever (Burhardt & Smiths, 1991); recent evidence also
1998; Saccheri et al., 2005; Asa et al., 2007). The reveals that inbreeding is an important risk factor in
inbreeding depression on male fertility in mammal susceptibility to infectious diseases such as tubercu-
species such as the Cuvier’s gazelles and Mexican losis and hepatitis (Lyons et al., 2009b). In addition,
grey wolves appears to be caused by an adverse effect an association between homozygosity and childhood
of inbreeding on sperm quality (Roldan et al., 1998; mortality resulting from invasive bacterial disease
Asa et al., 2007). has been also reported (Lyons et al., 2009a). The evi-
Our findings suggest that Charles Darwin’s sons dence of inbreeding depression on child survival in
probably experienced an adverse effect of inbreeding Charles Darwin’s offspring together with the findings
on fertility. Thus, the mean ± SE number of children reported in the present study of an adverse effect of
per woman was 1.250 ± 0.648 for Darwin’s five sons, inbreeding on fertility in his sons suggest that Dar-
whereas the mean ± SE family size for the non-inbred win’s fears on the health of his children as a result of
men was 2.100 ± 0.781 (Fig. 2). However, the family his marriage with his first cousin Emma Wedgwood
size was rather variable among Darwin’s sons. were neither unjustified nor exaggerated.
William and Leonard, who each married twice, had no
children; Francis had one child with each one of two
wives and no children with his third wife; and George
ACKNOWLEDGEMENTS
and Horace had five and three children, respectively.
The variation in family size among Darwin’s We thank Cesar Sánchez for help with the statisti-
sons may be partially attributed to the inherent cal analyses and Frank Nicholas for comments on
stochasticity of inbreeding. Thus, the proportion of an earlier version of the manuscript. We also thank
the autosomal genome expected to be homozygous by two anonymous reviewers for their helpful com-
descent in Charles Darwin’s children computed from ments. FCC and GA carried out the statistical
the genealogical information was 0.0630 with a rela- analysis. GA, FCC and TMB helped conceive the
tively large standard error of 0.0244 (see Material study, participated in its design and coordination,
and methods). Furthermore, a genome-wide scan and helped draft the manuscript. All authors read
based on 10 000 single nucleotide polymorphisms of and approved the final manuscript submitted for
individuals whose parents were first cousins showed publication. The authors declare that they have no
that those inbred individuals presented a variable conflict of interest.
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483482 G. ÁLVAREZ ET AL.
REFERENCES Freeman RB. 1982. The Darwin family. Biological Journal of
the Linnean Society 17: 9–21.
Álvarez G, Quinteiro C, Ceballos FC. 2011. Inbreeding and Freeman RB. 1984. Darwin pedigrees. London: Printed for
genetic disorder. In: Ikehara K, ed. Advances in the study of the Author.
genetic disorders. Croatia: In Tech, 21–44. Gibson JN, Morton NE, Collins A. 2006. Extended tracts of
Asa C, Miller P, Agnew M, Rebolledo JAR, Lindsey SL, homozygosity in outbred human populations. Human
Callahan M, Bauman K. 2007. Relationship of inbreeding Molecular Genetics 15: 789–795.
with sperm quality and reproductive success in Mexican Golubovsky M. 2008. Unexplained infertility in Charles
gray wolves. Animal Conservation 10: 326–331. Darwin’s family: genetic aspect. Human Reproduction 23:
Berra TM. 2013. Darwin & his children: his other legacy. 1237–1238.
New York, NY: Oxford University Press. Hayman J. 2013. Charles Darwin’s mitochondria. Genetics
Berra TM, Álvarez G, Ceballos FC. 2010a. Was the 194: 1–5.
Darwin/Wedgwood dynasty adversely affected by consan- Hedrick PW. 2011. Genetics of populations. Sudbury, MA:
guinity? BioScience 60: 376–383. Jones and Bartlett.
Berra TM, Álvarez G, Shannon K. 2010b. The Galton Helgason A, Pálsson S, Gudbjarstsson DF, Kristjánsson
–Darwin–Wedgwood pedigree of H. H. Laughlin. Biological P, Stefánsson K. 2008. An association between the
Journal of the Linnean Society 101: 228–241. kinship and fertility of human couples. Science 319: 813–
Bittles AH. 2009. Commentary: the background and outcomes 816.
of the first-cousin marriage controversy in Great Britain. Jones S. 2008. Darwin’s island: the Galapagos in the garden
International Journal of Epidemiology 38: 1453–1458. of England. London: Little Brown.
Bittles AH. 2012. Consanguinity in context. Cambridge: Cam- Keynes R. 2001. Annie’s box: Charles Darwin, his daughter
bridge University Press. and human evolution. London: Fourth State.
Bittles AH, Black ML. 2010. Consanguinity, human evolu- Kuper A. 2009. Commentary: a Darwin family concern.
tion, and complex diseases. Proceedings of the National International Journal of Epidemiology 38: 1439–1442.
Academy of Sciences of the United States of America 107: Lloyd-Smith JO. 2007. Maximum likelihood estimation
1779–1786. of the negative binomial dispersion parameter for highly
Bittles AH, Grant JC, Sullivan SG, Hussain R. 2002. Does overdispersed data, with applications to infectious diseases.
inbreeding lead to decreases human fertility? Annals of PLoS ONE 2: e180.
Human Biology 29: 111–130. Lyons EF, Amos W, Berkley JA, Mwangi I, Shafi M,
Browne J. 2002. Charles Darwin: the power of place. London: Williams TN, Newton CR, Peshu K, Marsh K, Scott
Jonathan Cape. JAG, Hill AVS. 2009a. Homozygosity and risk of childhood
Burhardt F, Smiths S. 1991. The correspondence of Charles death due to invasive bacterial disease. BMC Medical
Darwin, Vol. 6. Cambridge: Cambridge University Press. Genetics 10: 55.
Carothers AD, Rudan I, Kolcic I, Polasek O, Hayward C, Lyons EF, Frodsham AJ, Zhang L, Hill AVS, Amos W.
Wright AF, Campbell H, Teague P, Hastie ND, Weber 2009b. Consanguinity and susceptibility to infectious dis-
JL. 2006. Estimating human inbreeding coefficients: com- eases in humans. Biology Letters 5: 574–576.
parison of genealogical and marker heterozygosity McQuillan R, Leutenegge AL, Abdel-Rahman R,
approaches. Annals of Human Genetics 70: 666–676. Franklin CS, Pericic M, Barac-Lauc L, Smolej-
Cavalli-Sforza LL, Bodmer WF. 1971. The genetics of Narancic N, Janicijevic B, Polasek O, Tenesa A,
human populations. San Francisco, CA: WH Freeman. MacLeod AK, Farrington SM, Rudan P, Hayward C,
Ceballos FC, Álvarez G. 2013. Royal dynasties as human Vitart V, Rudan I, Wild SH, Dunlop MG, Wright AF,
inbreeding laboratories: the Habsburgs. Heredity 111: 114– Campbell H, Wilson F. 2008. Runs of homozygosity in
121. European populations. American Journal of Human Genet-
Colp R Jr. 2008. Darwin’s illness. Gainesville, FL: University ics 83: 359–372.
Press of Florida. Moore J. 2005. Good breeding: Darwin doubted his own
Darwin C. 1868. The variation of animals and plants under family’s ‘fitness. Natural History 114: 45–46.
domestication. London: John Murray. Ober C, Hyslop T, Hauck WW. 1999. Inbreeding effects on
Darwin C. 1876. The effects of cross and self fertilization in fertility in humans: evidence for reproductive compensation.
the vegetable kingdom. London: John Murray. American Journal of Human Genetics 64: 225–231.
Darwin GH. 1875. Marriages between first cousins in Pannell JR. 2009. On the problems of a closed marriage:
England and their effects. Fortnigh Rev 24: 22–41. celebrating Darwin 200. Biology Letters 5: 332–335.
Reprinted in Int J Epidemiol 2009, 38:1429–1439. Postma E, Martini L, Martini P. 2010. Inbred women in a
Falconer DS, Mackay TFC. 1996. Introduction to quantita- small Swiss village have fewer children. Journal of Evolu-
tive genetics. Harlow: Longman Group. tionary Biology 23: 1468–1474.
Fenner L, Egger M, Gagneux S. 2009. Annie’s death, the R Development Core Team: R. 2011. A language and envi-
evolution of tuberculosis and the need for systems epidemi- ronment for statistical computing. Vienna: R Foundation
ology. International Journal of Epidemiology 38: 1425– for Statistical Computing, Available at: http://www.R
1428. -project.org
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483INBREEDING DEPRESSION IN THE DARWIN FAMILY 483
Robert A, Toupance B, Tremblay M, Heyer E. 2009. data using the negative binomial distribution. Ecology 77:
Impact of inbreeding on fertility in a pre-industrial popula- 2549–2557.
tion. European Journal of Human Genetics 17: 673–681. Woods CG, Cox J, Springell K, Hampshire DJ, Mohamed
Roldan ERS, Cassinello J, Abaigar T, Gomendio M. MD, McKibbin M, Raymond FL, Sanford R, Sharif SM,
1998. Inbreeding, fluctuating asymmetry and ejaculate Karbani G, Ahmed M, Bond J, Clayton D, Inglehearn
quality in an endangered ungulate. Proceedings of the Royal CF. 2006. Quantification of homozygosity in consanguineous
Society of London Series B, Biological Sciences 265: 243– individuals with autosomal recessive disease. American
248. Journal of Human Genetics 78: 889–896.
Saccheri IJ, Lloyd HD, Helyar SJ, Brakefield PM. 2005. Zeileis A, Kleibe C, Jackman S. 2008. Regression models
Inbreeding uncovers fundamental differences in the genetic for count data in R. Journal of Statistical Software 27:
load affecting male and female fertility in a butterfly. Pro- 1–25.
ceedings of the Royal Society of London Series B, Biological Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM.
Sciences 272: 39–46. 2009. Mixed effects models and extensions in ecology with R.
White GC, Bennetts RE. 1996. Analysis of frequency count New York, NY: Springer Science.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher’s web-site:
Table S1. Regression analyses of the number of children per woman. The table reports the results correspond-
ing to the four regression models considered in the article: Zero-inflated Poisson (ZIP) and zero-inflated negative
binomial (ZINB), Poisson generalized linear model (GLM) and negative binomial GLM.
Table S2. Regression analyses of protogenesic and intergenesic intervals. The table reports the results
corresponding to regression analyses of protogenesic and intergenesic intervals by ordinary least-squares.
Table S3. Regression analyses of the reproductive span. The table reports the results corresponding to the four
regression models considered in the article: Zero-inflated Poisson (ZIP) and zero-inflated negative binomial
(ZINB), Poisson generalized linear model (GLM) and negative binomial GLM.
© 2014 The Linnean Society of London, Biological Journal of the Linnean Society, 2015, 114, 474–483You can also read
