ORIGINAL PAPER
Niger-wide assessment of in situ sorghum genetic diversity
with microsatellite markers
M. Deu Æ F. Sagnard Æ J. Chantereau Æ C. Calatayud Æ D. He
´
rault Æ
C. Mariac Æ J.-L. Pham Æ Y. Vigouroux Æ I. Kapran Æ P. S. Traore Æ
A. Mamadou Æ B. Gerard Æ J. Ndjeunga Æ G. Bezanc¸on
Received: 4 September 2007 / Accepted: 23 January 2008 / Published online: 14 February 2008
Springer-Verlag 2008
Abstract Understanding the geographical, environmental
and social patterns of genetic diversity on different spatial
scales is key to the sustainable in situ management of genetic
resources. However, few surveys have been conducted on
crop genetic diversity using exhaustive in situ germplasm
collections on a country scale and such data are missing for
sorghum in sub-Saharan Africa, its centre of origin. We
report here a genetic analysis of 484 sorghum varieties col-
lected in 79 villages evenly distributed across Niger, using 28
microsatellite markers. We found a high level of SSR
diversity in Niger. Diversity varied between eastern and
western Niger, and allelic richness was lower in the eastern
part of the country. Genetic differentiation between botani-
cal races was the first structuring factor (Fst = 0.19), but the
geographical distribution and the ethnic group to which
farmers belonged were also significantly associated with
genetic diversity partitioning. Gene pools are poorly differ-
entiated among climatic zones. The geographical situation of
Niger, where typical western African (guinea), central
African (caudatum) and eastern Sahelian African (durra)
sorghum races converge, explained the high observed
genetic diversity and was responsible for the interactions
among the ethnic, geographical and botanical structure
revealed in our study. After correcting for the structure of
botanical races, spatial correlation of genetic diversity was
still detected within 100 km, which may hint at limited seed
exchanges between farmers. Sorghum domestication his-
tory, in relation to the spatial organisation of human
societies, is therefore key information for sorghum in situ
conservation programs in sub-Saharan Africa.
Introduction
Characterizing the patterns of in situ crop diversity and
understanding the underlying evolutionary processes that
have shaped the observed genetic structures are two pre-
requisites for both breeding and plant genetic resources
programs. This is especially important for traditional
cereals cultivated in their centre of diversity, where unique
Communicated by F. Ordon.
Electronic supplementary material The online version of this
article (doi:10.1007/s00122-008-0721-7) contains supplementary
material, which is available to authorized users.
M. Deu (&) F. Sagnard C. Calatayud D. He
´
rault
CIRAD, UMR DAP, Avenue Agropolis, TA A 96/03,
34398 Montpellier, France
e-mail: monique.deu@cirad.fr
F. Sagnard P. S. Traore
International Crop Research Institute for the Semi-Arid Tropics
(ICRISAT), Station de Samanko, BP320, Bamako, Mali
J. Chantereau
CIRAD, UPR Agrobiodiversite
´
en savane,
34398 Montpellier, France
C. Mariac J.-L.Pham
Institut de Recherche pour le De
´
veloppement (IRD),
Montpellier, France
Y. Vigouroux G. Bezanc¸on
Institut de Recherche pour le De
´
veloppement (IRD),
Niamey, Niger
I. Kapran A. Mamadou
Institut National de la Recherche Agronomique du Niger
(INRAN), Niamey, Niger
B. Gerard J. Ndjeunga
International Crop Research Institute for the Semi-Arid Tropics
(ICRISAT), Niamey, Niger
123
Theor Appl Genet (2008) 116:903–913
DOI 10.1007/s00122-008-0721-7
genes or gene complexes can be found and often represent
valuable genetic resources for breeders (Brush 1995;
Maxted et al. 2002). In decentralised breeding programs,
determining the genetic structure of local landraces, which
are often preferred by farmers for their adaptation, taste or
post-harvest processing traits, should help choose the best
entries and delineate target release zones for improved
varieties (Ceccarelli et al. 1997). The geographical distri-
bution of landrace vernacular names, agromorphological
types and genetic diversity on different spatial scales also
provides valuable information to complement ex situ col-
lections, and establish relevant criteria to initiate and
monitor in situ conservation programs (Brush 2000).
Sorghum (Sorghum bicolor L. Moench) was domesti-
cated in northeastern Africa. It is an annual, predominantly
selfing cereal (Ollitrault et al. 1997; Dje
`
et al. 2004). Cul-
tivated sorghums (Sorghum bicolor ssp. bicolor) have been
classified into five basic botanical races (bicolor, caudatum,
durra, guinea and kafir) and ten intermediate ones, based on
panicle and spikelet morphology (Harlan and de Wet 1972).
Most sorghum genetic surveys relied on gene bank
accessions to assess large-scale geographical or taxonomic
structures, either with allozyme or DNA markers (Aldrich
et al. 1992; Deu et al. 1994, 1995, 2006; Cui et al. 1995;de
Oliveira et al. 1996; Menkir et al. 1997; Dje
`
et al. 2000;
Grenier et al. 2000; Casa et al. 2005). Fewer studies have
involved in situ collection of sorghum landraces and
detailed information on locations, growing environments
and farmers’ practices. Surveys based on in situ collections
generally reveal a lack of correlation between genetic
diversity parameters and environmental factors (Ayana
et al. 2000, 2001; Ghebru et al. 2002; Zongo et al. 2005)
and weak genetic differentiation between regions (Ollitra-
ult et al. 1997; Dje
`
et al. 1999; Nkongolo and Nsapato
2003; Kayode
´
et al. 2006). As they are based on a relatively
small number of villages that are not exhaustively sampled,
limited inferences can be drawn as to the evolutionary
factors responsible for the observed genetic structures.
In Niger, sorghum is the second most cultivated crop
after pearl millet. It is grown under rainfed conditions in
traditional farming systems. Water availability is the main
limiting factor for agriculture. Rainfall varies from year to
year and over short distances. A desiccation process has
been observed especially in the dryer regions of the
country, and Hulme (2001) evaluated that annual rainfall
across the Sahel decreased by 20–30% between the 1930–
1950 and the 1970–1990 periods. Over the past three
decades, annual sorghum production has steadily increased,
in response to the doubling of the population. This involved
an extension of sorghum cultivation to marginal lands and
an impoverishment of the soils due to shortened fallows
(Wezel and Boecker 1998). Three major ethnic groups
practice agriculture and small-scale plant breeding. The
Zarma/Songhaı
¨
(22% of the population of Niger), the
Hausa (56%) and the Kanuri (4.3%) are predominant in
western, central and eastern regions, respectively (Fig. 1).
Adoption of improved varieties is limited in Niger, and
the largest part of sorghum production comes from farmer-
selected landraces. These landraces are locally adapted to
their harsh, heterogeneous and unpredictable environments.
Ethnic traditions, social organizations and food preferences
also probably contribute to the extent and structure of crop
diversity (Reenberg 2001). In situ conservation programs
therefore require the identification and understanding of the
drivers of genetic diversity dynamics.
The objectives of this study were to (1) characterize
genetic diversity in a sorghum collection from 79 villages
covering the complete geographical and agro-ecological
range of sorghum growing areas in Niger, (2) assess the
patterns of genetic diversity revealed by SSR markers in
relation to botanical, climatic and ethnic factors, (3) measure
the spatial structure of genetic diversity on different scales
and (4) propose evolutionary scenarios of sorghum diversity
in Niger as a basis for future in situ conservation programs.
Material and methods
Collection of sorghum varieties
From October 2003 to December 2003, an intensive col-
lection of sorghum varieties was conducted in 79 villages
covering the rainfall gradient and range of agro-ecological
conditions of Niger’s agricultural areas (Fig. 1). A multi-
disciplinary team composed of social scientists, agrono-
mists and geneticists, assisted by Zarma and Hausa
translators, collected seeds and interviewed farmers in two
villages per day on average. Varietal inventory also
included information on crop uses, seed origins, agricul-
tural systems and social organizations in the villages. In
each village, we sampled all local varieties listed by a
representative group of farmers. Each variety was provided
by one farmer in one village, either by grains (unselected
panicles) or by seeds. The varieties were collected either in
farmers’ fields, in panicle-drying zones or in home grana-
ries, depending on the advancement of harvest work. For
each variety, a bulk of seeds was collected from 30 pani-
cles. We assume that we sampled the majority of sorghum
varieties grown in these 79 villages in 2003.
Racial characterisation of sorghum varieties
The whole collection was grown in 2004 at INRAN
(Institut National de la Recherche Agronomique du Niger)
experimental stations in Maradi and Bengou.
904 Theor Appl Genet (2008) 116:903–913
123
Racial characterisation based on panicle and spikelet
morphology was carried out in accordance with Harlan and
de Wet’s (1972) classification. In addition, Snowden’s
taxon margaritiferum was distinguished, since those small-
grain guinea sorghums appear to be genetically different
from all other representatives of the guinea race (Folk-
ertsma et al. 2005; Deu et al. 2006).
DNA extraction and SSR genotyping
Sorghum seeds were germinated in a greenhouse. For
each variety, DNA was isolated from fresh leaves col-
lected on one 2- to 3-week-old seedling following a
CTAB protocol described by Deu et al. (1995). Hereafter,
the genetic diversity of a variety refers to the multilocus
genotype of one single plant representative of a maternal
parent (i.e. landrace or improved variety). The genetic
information provided by one single individual per variety,
with no information on within-variety diversity, has
proved to be sufficient to detect large-scale inter-varietal
evolutionary trends even in outcrossing crops when the
number of loci is sufficient (Matsuoka et al. 2002; Mariac
et al. 2006).
Twenty-eight SSR markers were assayed (listed as elec-
tronic supplementary information S1). They formed a subset
of 50 microsatellites selected for their reliability and scoring
accuracy between laboratories, and their level of polymor-
phism and genome coverage for the Generation Challenge
Program (http://sat.cirad.fr/sat/sorghum_SSR_kit). Most of
these markers have been previously described and mapped:
Sb (Brown et al. 1996; Taramino et al. 1997), Xcup (Schloss
et al. 2002), Xtxp (Bhattramakki et al. 2000; Menz et al.
2002). Gpsb markers were developed in CIRAD and mapped
in our RILs population (to be published elsewhere). PCR
conditions and genotyping on Li-Cor automated sequencers
were as described by Barnaud et al. (2007). Saga GT v. 2.2
(Li-Cor) was used to determine allele sizes. Genotyping was
conducted at the Languedoc Roussillon Genotyping Plat-
form hosted by CIRAD.
Investigated factors of genetic structure
The partitioning of microsatellite diversity was explored on
different scales according to racial (botanical), eco-geo-
graphical, climatic, ethnic and seed origin criteria using
information provided by the collection questionnaires or
available in meteorological databases. Three longitudinal
classes (4 interval) were created. They were roughly
aligned on the administrative departments of Tillaberi and
Dosso (Western), Tahoua and Maradi (Central) and Zinder
and Diffa (Eastern). Three main annual rainfall classes
were defined: ‘‘less than 400 mm’’, ‘‘from 400 to 500 mm’’
and ‘‘more than 500 mm’’. Varieties were also clustered
according to the dominant ethnic groups in the villages
(Hausa, Kanuri and Zarma/Songhaı
¨
) and their date of
introduction (recent or ancient).
Fig. 1 Location of the 79
villages visited for the
collection of 484 sorghum
varieties in Niger. Isohyets were
computed using ArcGIS
geostatistical analyst (v. 9.0)
based on 1971–2000 annual
rainfall normals (linear kriging
with smoothing). The backdrop
ethnic map is a simplified
distribution of dominant ethnic
groups adapted from linguistic
maps of Africa (http://www.
muturzikin.com, 2007)
Theor Appl Genet (2008) 116:903–913 905
123
Genetic data analyses
Genetic diversity parameters were estimated for each
defined group of varieties with GENETIX software 4.04
(Belkhir et al. 2002): total number of alleles (A
t
), number
of rare alleles (A
r
, freq \ 5%), observed heterozygosity
(H
o
) and expected heterozygosity or gene diversity (H
e
)
adjusted for low sampling size, according to Nei (1978),
following formula
H ¼ 2n 1
X
j
X
i
x
2
ij
!,
r
"#,
2n 1ðÞ
where x
ij
is the frequency of the ith allele of locus j, r is the
number of genetic loci and n is the population size. We also
calculated R
s
(allelic richness) using the rarefaction method
(Petit et al. 1998) implemented in FSTAT (Goudet 2001).
This method permits to estimate the expected number of
different alleles among equal-sized samples, based on the
lowest sample size in groups of varieties that are defined
for comparisons. Consequently, allelic richness was
calculated for 16 varieties (32 genes) for botanical
comparisons corresponding to the smallest size of the
guinea margaritiferum sample but for 86 varieties (172
genes) for rainfall classes. The expectation of the number
of alleles in a sample size of g (g B N)is
^
r
ðgÞ
¼
X
i
1
C
g
NN
i
C
g
N
where a
1
,…,a
k
are the different alleles at a single locus and
N
1
,…,N
k
are the number of times they appeared in the N-
size sample of each group.
The significance of differences in R
s
and H
e
between the
defined groups was tested using a Wilcoxon signed-rank
test across loci.
To investigate the genetic relationships between varie-
ties, a genetic dissimilarity matrix was computed using the
simple matching index. A dendrogram was then generated
on the dissimilarity matrix with the Neighbour Joining (NJ)
algorithm implemented in DARWIN V5 software (Perrier
et al. 2003). Only one variety with too many missing data
was not included in the cluster analysis.
We explored the population structure using Weir and
Cockerham’s (1984) h as an estimate of Fst
.
Overall Fst
and pairwise Fst were calculated and tested for their sig-
nificance with GENETIX software 4.04. Fst was only
interpreted as a descriptive differentiation parameter in our
study, since the mixing of different varieties in the same
group prevented us from drawing any evolutionary infer-
ences from the estimated F-statistics.
However, analysis of the correlation between the
genetic relatedness among varieties and the geographical
distances provided indirect insight into several
evolutionary factors, including limited gene flow respon-
sible for isolation-by-distance processes (Sokal and
Wartenberg 1983; Smouse and Peakall 1999; Rousset
2000). Under restricted gene flow through either pollen or
seed movements, individuals or varieties growing close to
each other are more genetically similar than those grow-
ing far apart. The mean pairwise genetic kinship among
varieties within short distances is then higher than the
average kinship between all sampled individuals. For
spatial genetic structure analyses, distance classes were
determined to ensure that a sufficient number of data were
included in the computations for each distance interval.
We used the individual kinship estimator proposed by
Ritland (1996) averaged across loci as a measurement of
genetic similarity between varieties. The statistical sig-
nificance of kinship estimators was tested using 1,000
Monte Carlo simulations on the hypothesis of complete
spatial randomness of genetic diversity. We first con-
ducted the spatial analyses on the country-scale
collection, and secondly, within three genetic sub-groups
identified from the NJ dendrogram to assess fine scale
patterns after correcting for the larger racial structure. All
the spatial computations were performed using SPAGeDi
(ver 1.2) software (Hardy et al. 2002).
Results
Racial sorghum diversity in Niger
The collection studied here comprises 484 sorghum vari-
eties obtained in 79 villages (Fig. 1). The number of
varieties collected per village ranges from zero to 16
(average: 6.13). Varietal richness varies across regions with
6.23, 8.25 and 3.88 varieties per village, in western, central
and eastern Niger, respectively. No difference is observed
between the two most important ethnic groups with 6.39
and 6.05 varieties per village in 48 Hausa and 18 Zarma/
Songhaı
¨
villages, respectively.
All sorghum basic botanical races except kafir can be
found in Niger (listed as electronic supplementary infor-
mation S2). In this collection, durra (23.1%) and
caudatum (21.3%) races are the most prevalent, followed
by guinea (10.5%) and bicolor (8.1%). Among guinea,
43.1% could be identified as guinea margaritiferum vari-
eties. Intermediate races (25.8%) between durra and
bicolor (49 varieties), durra and caudatum (35 varieties),
caudatum and bicolor (19 varieties) and caudatum and
guinea (17 varieties) amount to a significant share of the
collection. About 11% of the varieties could not be
classified either because entire panicles were missing or
because they were hybrids with a complex racial
pedigree.
906 Theor Appl Genet (2008) 116:903–913
123
Racial distribution is not random between regions.
Guinea varieties are mainly cultivated in western Niger,
and principally in the southernmost areas where annual
rainfall is more abundant. Caudatum and bicolor varieties
are rare in eastern Niger. Durra varieties cover all the
sorghum-growing areas, but are predominant in the eastern
region. A high racial diversity is observed in central Niger,
though caudatum and intermediate races prevail.
Overall genetic diversity
Twelve varieties were discarded from the SSR analysis due
to their weak germination or insufficient yield during DNA
extractions. Overall, we present the analysis of 472 varie-
ties from 76 villages. The average missing data per
accession was 2.6%.
The 28 microsatellite loci are polymorphic and reveal
292 alleles (listed as electronic supplementary information
S1), from 2 to 26 alleles per locus with an average of 10.43.
Sixty-four percent of alleles are rare at the 0.05 threshold.
Observed heterozygosity ranges from 1.7% (gpsb067) to
7.8% (Xtxp15) with an average of 4.2%. The 28 SSR loci
can discriminate 454 varieties out of 472 (96%).
Genetic diversity estimates
As expected, the allelic richness is highly sensitive to the
rare alleles frequently revealed by SSR markers. This
parameter increases with the number of sampled genes,
which varies among groups that are compared. R
s
is
therefore only meaningful to assess the genetic diversity
differences that occur within one categorical factor (i.e.
between races for the racial structuration or between rain-
fall classes for the climatic groups).
Allelic richness is significantly higher in the caudatum
(R
s
= 4.88 alleles per locus) race than in durra (3.94;
P \ 0.01), guinea (4.11; P \ 0.05) and guinea margari-
tiferum (3.21; P \ 0.01) (Table 1). Within the guinea race,
the margaritiferum group has significantly lower allelic
richness (P \ 0.05). The gene diversity estimates exhibit
the same trends except that the durra race (H
e
= 0.393)
appears to be less genetically diverse than bicolor (0.571;
P \ 0.001) and guinea (0.496; P \ 0.05). Overall Fst is
high among the basic races (Fst = 0.278, P \ 0.01) but
remains substantial when intermediate races are also taken
into account (Fst = 0.19, P \ 0.01). Pairwise Fst are all
significant at the 1% threshold and illustrate particularly
high genetic differentiation between the margaritiferum
Table 1 Genetic diversity
estimates between structuring
factors
N number of landraces included
in each group, A
t
total number
of alleles, A
p
number of private
alleles (present in a single
group), R
s
Allelic richness for
each group based on a minimum
sample of gene copies (the
minimum sample is indicated in
brackets), H
e
gene diversity, H
o
observed heterozygosity, Fst
average genetic differentiation
for each group
a
For ‘‘racial’’ Fst, a second
estimation of Fst has been
calculated for basic races only
on a population of 299 landraces
after exclusion of intermediate
and unclassified landraces
NA
t
A
p
R
s
H
e
H
o
Fst
Races 472(299
a
) 5.50(32) 0.190(0.278
a
)
Bicolor 36 144 9 4.43 0.571 0.038
Caudatum 101 192 16 4.88 0.569 0.039
Durra 111 174 17 3.94 0.393 0.029
Guinea 29 130 8 4.11 0.496 0.071
Guinea m. 22 92 7 3.21 0.427 0.041
Intermediate 120 195 11 4.73 0.545 0.050
Unclassified 53 160 4 4.65 0.582 0.044
Regions 472 8.46(172) 0.070
West 186 239 38 7.92 0.646 0.060
Centre 193 233 27 7.80 0.603 0.030
East 93 172 12 6.12 0.422 0.033
Rainfall classes 472 8.14(142) 0.031
\400 mm 246 253 32 7.70 0.580 0.040
400–500 mm 145 235 17 7.87 0.624 0.035
[500 mm 81 195 16 6.93 0.622 0.060
Ethnic groups 459 6.21(46) 0.054
Hausa 307 257 51 5.96 0.590 0.032
Zarma/Songhaı
¨
124 226 27 6.22 0.650 0.068
Kanuri 28 112 5 3.93 0.393 0.044
Origins 428 8.94(248) 0.008
Recent introduction 136 221 18 7.88 0.586 0.034
Ancient 292 269 65 8.75 0.618 0.046
Total collection 472 292 10.43 0.613 0.042
Theor Appl Genet (2008) 116:903–913 907
123
