PAX4 gene variations predispose to ketosis-prone
Franck Mauvais-Jarvis1,3,4,*, Stuart B. Smith3, Ce ´dric Le May1, Suzanne M. Leal2,
Jean-Franc ¸ois Gautier4, Mariam Molokhia5, Jean-Pierre Riveline6, Arun S. Rajan1,
Jean-Philippe Kevorkian7, Sumei Zhang3, Patrick Vexiau4, Michael S. German3
and Christian Vaisse3
1Division of Diabetes, Endocrinology and Metabolism, Department of Medicine and2Department of Molecular and
Human Genetics, Baylor College of Medicine, Houston, TX 77030, USA,3Diabetes Center, Department of Medicine,
University of California, San Francisco, San Francisco, CA 94143, USA,4Department of Endocrinology and Diabetes,
Saint-Louis Hospital, University of Paris VII School of Medicine, Paris 75010, France,5Epidemiology Unit, London
School of Hygiene and Tropical Medicine, London, WC1E 7HT, UK,6Department of Diabetes and Metabolic Diseases,
Sud Francilien Hospital, Corbeil-Essonnes 91100, France and7Department of Internal Medicine B, Lariboisiere
Hospital, Paris, 75010, France
Received August 4, 2004; Revised and Accepted October 15, 2004
Ketosis-prone diabetes (KPD) is a rare form of type 2 diabetes, mostly observed in subjects of west African
origin (west Africans and African-Americans), characterized by fulminant and phasic insulin dependence, but
lacking markers of autoimmunity observed in type 1 diabetes. PAX4 is a transcription factor essential for the
development of insulin-producing pancreatic b-cells. Recently, a missense mutation (Arg121Trp) of PAX4
has been implicated in early and insulin deficient type 2 diabetes in Japanese subjects. The phenotype simi-
larities between KPD and Japanese carriers of Arg121Trp have prompted us to investigate the role of PAX4 in
KPD. We have screened 101 KPD subjects and we have found a new variant in the PAX4 gene (Arg133Trp),
specific to the population of west African ancestry, and which predisposes to KPD under a recessive model.
Homozygous Arg133Trp PAX4 carriers were found in 4% of subjects with KPD but not in 355 controls or 147
subjects with common type 2 or type 1 diabetes. In vitro, the Arg133Trp variant showed a decreased tran-
scriptional repression of target gene promoters in an alpha-TC1.6 cell line. In addition, one KPD patient
was heterozygous for a rare PAX4 variant (Arg37Trp) that was not found in controls and that showed a
more severe biochemical phenotype than Arg133Trp. Clinical investigation of the homozygous Arg133Trp
carriers and of the Arg37Trp carrier demonstrated a more severe alteration in insulin secretory reserve,
during a glucagon-stimulation test, compared to other KPD subjects. Together these data provide the first
evidence that ethnic-specific gene variants may contribute to the predisposition to this particular form of dia-
betes and suggest that KPD, like maturity onset diabetes of the young, is a rare, phenotypically defined but
genetically heterogeneous form of type 2 diabetes.
Rare genetic defects in transcription factors controlling the
expression of the insulin gene and other proteins critical for
normal pancreatic b-cell metabolism/function impair insulin
secretion and lead to diabetic syndromes characterized by a
rapid evolution toward insulin deficiency, such as maturity
onset diabetes of the young (MODY) (1–5). PAX4 is a tran-
scription factor that plays a critical role in the differentiation
of embryonic pancreatic progenitors into insulin-producing
b-cells. Mice with targeted disruption of the pax4 gene
show absence of mature insulin-producing b-cells, and die
in the first days of life from severe insulin deficient diabetes
(6). Although PAX4 gene mutations have not been found in
Human Molecular Genetics, Vol. 13, No. 24 # Oxford University Press 2004; all rights reserved
*To whom correspondence should be addressed at: Division of Diabetes, Endocrinology and Metabolism, Department of Medicine, Baylor College of
Medicine, One Baylor Plaza, BCMA 700B, Houston, TX 77030, USA. Tel: þ1 7137987224; Fax: þ1 7137983810; Email: firstname.lastname@example.org
Human Molecular Genetics, 2004, Vol. 13, No. 24
Advance Access published on October 27, 2004
MODY families, a rare missense mutation in PAX4
(Arg121Trp) has recently been described in the Japanese
type 2 diabetic population. Japanese carriers of the Arg121Trp
variant are characterized by either a transient insulin depen-
dence at diabetes onset or a rapid evolution toward insulin
deficiency, suggesting that PAX4 mutations lead to severe
b-cell dysfunction in humans (7).
Ketosis-prone diabetes (KPD) belongs to a rare subgroup of
type 2 diabetes with severe insulin deficiency, mostly observed
in subjects of sub-Saharan African ancestry, such as west
Africans, Caribbeans and African-Americans (8–14). Its
phenotype is distinct from the common type 2 and type 1
diabetes. It is characterized by a fulminant initial insulin
dependence, without the immunological markers observed in
classical type 1 diabetes, followed by a subsequent clinical
course which varies from non-insulin treated type 2 diabetes
to insulin-dependent idiopathic type 1 diabetes (14). A
severe dysfunction of the insulin-producing b-cells is attested
to by the observation that 25% of the subjects are insulin
dependent at diabetes onset, whereas the remaining 75% will
develop permanent insulin dependence within 10 years (14).
The clinical similarities between Japanese type 2 diabetic
carriers of the Arg121Trp PAX4 mutation and west African
subjects with KPD suggest that mutations in PAX4 could
also predispose to the latter.
We studied a cohort of west African subjects with KPD for
mutations in the PAX4 gene. Our results provide the first evi-
dence that an ethnic-specific variant of PAX4 (Arg133Trp),
associated with functional alterations in vivo and in vitro,
predisposes to KPD, and represents a marker of severe
insulin deficiency in the population of west African descent.
Screening of west African ketosis-prone diabetic
subjects for variants in PAX4
We sequenced the coding region of the PAX4 gene in 101 unre-
in Table 1. We found two new missense variants, one common
variant already described, and three silent polymorphisms.
Four subjects (4%) carried a new homozygous missense
variant resulting in the conversion of arginine (R) to trypto-
phane (W) at position 133 (R133W). The heterozygous
R133W variant was found in 27 subjects. One subject (1%)
carried a new missense mutation R37W. The P321H variant
has already been described in the Caucasian and Japanese
populations (15). The P321H and the three silent variants
were not associated with KPD in our population (Table 1).
As was expected, owing to their close physical proximity
there was a strong linkage disequilibrium (P , 0.00001)
between variants R133W and P321H in both the control
(D0¼ 0.97) and ketosis-prone diabetic subjects (D0¼ 0.82).
Most patients studied are immigrants from west African
countries and their first and second degree relatives were not
The R133W variant is not found in Caucasians
To investigate the ethnic specificity of the R133W variant, we
evaluated its prevalence by sample genotyping a Caucasian
control population. In this population, PAX4 was monomor-
phic at the 133R site and the 133W allele was not found in
a total of 200 individuals (Fig. 2).
R133W is found at the homozygous state in KPD only
To evaluate the specificity of the R133W variant to KPD, we
genotyped an extended population of diabetic and controls of
west African origin (see Materials and Methods) (Table 2). All
groups were in Hardy–Weinberg equilibrium.
The homozygous R133W variant was found in ketosis-
prone diabetic subjects only (4%), and was not found in 147
type 2 diabetic, type 1 diabetic and 355 controls of west
African origin (Table 2). The frequency of homozygous
R133W carriers was compared between ketosis-prone diabetic
subjects and west African non-diabetic controls by separating
homozygous (W/W) individuals and wild-type (R/R)/hetero-
zygous (R/W) individuals into two different groups. Under
these conditions, the homozygous R133W carriers were
associated with an increased risk of KPD with an odds ratio
(OR) of 23.5 [95% confidence interval (CI), 2.7–203.6,
P ¼ 0.001].
The frequency of the heterozygous R133W variant was
increased in ketosis-prone diabetic subjects compared with
common type 2 diabetic subjects (x2¼ 4.91; P ¼ 0.026)
(Table 2). The increased frequency of heterozygous R133W
in ketosis-prone diabetic subjects did not reach significance
(x2¼ 1.54, P ¼ 0.21), owing to small sample size, but
reached significance when compared with African-Americans
(x2¼ 8.04, P ¼ 0.04). However, in this latter group, the
genetic admixture with Caucasians certainly lowers the
133W allele frequency. In any case, the 133W allele frequency
was higher in ketosis-prone diabetic subjects than in other
west African diabetic and control groups (Table 2) with an
OR .1.5, suggesting that the heterozygous R133W variant
may also increase the risk of KPD.
Interaction between R133W and P321H
As R133W and P321H are both frequent variants from the
same gene, we assessed whether the 321H allele could
modify the diabetic risk associated with the 133W allele by
comparing the haplotypes frequencies associated with these
alleles between ketosis-prone diabetic subjects and controls
(Table 3). The haplotypes frequencies were different in sub-
jects and controls (x2¼ 9.74; P ¼ 0.02), further suggesting
that the PAX4 locus predisposes to KPD. As expected, when
comparing the haplotypes R133/P321 and R133/321H (effect
of the P321H variant alone) there was no increased risk of dia-
betes [OR: 0.97 (CI: 0.63–1.47); P ¼ 0.85]. However, when
comparing the haplotypes 133W/321H to 133W/P321 (effect
of P321H on R133W), we observed a trend toward an
133W/P321[OR:4.48(CI:0.82–24.32),P ¼ 0.079].Together,
these data suggest that the 321H allele may partially reverse
the diabetes risk associate with the 133W allele (Table 3).
Owing to the small sample size we cannot draw any
3152Human Molecular Genetics, 2004, Vol. 13, No. 24
PAX4 variants show diminished ability to repress
PAX4 is expressed in the early embryonic pancreas and is not
expressed in adult islets (16). It is required to specify and
maintain the islet b-cell fate while repressing non-b-cell
fate. PAX4 is a transcription factor that represses gene
expression through a paired domain and homeodomain (16).
Figure 1A shows the position of the R37W and R133W
mutations in the PAX4 molecule. We assessed the functional
significance of the R37W and R133W variants by looking at
their ability to repress gene expression, following transfection
in a-TC1.6 cells. We chose the a-cell because of previous
work demonstrating that PAX4 repression of target promoters
in vitro is most fully achieved in a-cells, whereas it is weak
in b-cells (16,17). Similarly, we chose the insulin promoter
in our functional assay because wild-type PAX4 binds to it
with a higher affinity than other promoters tested (16).
a-TC1.6 cells were co-transfected with an insulin promoter
reporter construct along with each human PAX4 variant
(Fig. 1B). Wild-type PAX4 represses promoter activity by
70% compared with the control (compare lane 3 with lane 2).
On the other hand, both PAX4 variants R37W and R133W
repress reporter activity by only 45 and 37% of control,
respectively (compare lanes 4 and 5 with lane 2).
PAX4 (R37W) binds target genes with lower affinity
than wild-type PAX4
The ability of a transcription factor to stimulate or inhibit gene
expression is partly secondary to its ability to bind DNA
Table 1. Variants of the PAX4 gene in ketosis-prone diabetes
Nucleotide changeAmino acid changeDesignation Genotype frequency
Ketosis-prone diabetic subjectsControls
C109T Arg . TrpR37W C/T: 1 (1%)
C/C: 100 (99%)
C397T Arg . TrpR133W T/T: 4 (4%)
C/T: 27 (26.7%)
C/C: 70 (69.3%)
See Table 2
C450T Gly . Gly G150G C/T: 2 (2%)
C/C: 99 (98%)
A519G Gln . GlnQ173QA/A: 74 (73%)
G/A: 21 (21%)
G/G: 6 (6%)
C612T Asp . Asp D204DC/T: 1 (1%)
T/T: 100 (99%)
C962A Pro . His P321H A/A: 10 (10%)
C/A: 44 (44%)
C/C: 47 (46%)
aDifferences in genotype frequencies between ketosis-prone diabetic subjects and controls of west African origin were tested using the x2test.
Table 2. Frequency of the R133W variant of PAX4 in west African populations
R/RR/W W/W Frequency (95% CI)OR (95% CI)a
West African diabetic subjects
Ketosis-prone diabetesc(n ¼ 101)
Type 2 diabetesc(n ¼ 106)
Type 1 diabetesc(n ¼ 41)
0.173 (0.124, 0.233)
0.076 (0.044, 0.120)
2.57 (1.37, 4.66)
x2¼ 9.14, P ¼ 0.0025
1.94 (0.85, 4.05)
x2¼ 2.59, P ¼ 0.108
33 (82.9%) 8 (19.5%)0 (0%)0.110 (0.052, 0.198)
West African controls
Non-diabetic west African controls (n ¼ 255)
African-American controls (n ¼ 100)
Caucasian controls (n ¼ 200)
200 (78.4%) 55 (21.6%)0 (0%)0.106 (0.060, 0.162)1.77 (1.12, 2.79)
x2¼ 5.90, P ¼ 0.015
2.72 (1.44, 5.16)
x2¼ 9.99, P ¼ 0.0016
86 (86%)14 (14%)0 (0%)0.070 (0.039, 0.115)
200 (100%)0 (0%)0 (0%)–
aOdds ratios and confidence intervals evaluate the calculated risk of KPD when comparing the frequency of the R133W allele between the ketosis-
prone diabetic subjects and each diabetic and control groups.
bx2test was used to evaluate if the OR in ketosis-prone diabetic subjects compared with each diabetic and control group was significant.
cKetosis-prone diabetic, diabetic and type 1 diabetic subjects are west African natives.
Human Molecular Genetics, 2004, Vol. 13, No. 243153
sequence on the target gene promoter. We studied the binding
of R133W and R37W to the insulin promoter and to the PAX4
promoter in EMSA conditions (Fig. 1C). Both wild-type
PAX4 and R133W mutants showed similar binding activities
to specific DNA sequences. On the other hand, the binding
of the R37W variant to DNA sequences was decreased by
50% of that of the wild-type PAX4.
Clinical features of the homozygous R133W and
R37W PAX4 variant carriers
Table 4 shows the individual clinical and metabolic features of
ketosis-prone diabetic patients 1–4 carrying the homozygous
R133W variant and patient 5 carrying the R37W variant.
All patients originated from west African countries, and all
had a family history of diabetes, except one for which family
history was unknown. All patients showed very low C-peptide
levels and major hyperglucagonemia despite hyperglycemia at
initial admission, consistent with severely altered pancreatic
a- and b-cell function during diabetic ketoacidosis.
Among homozygous R133W carriers, patients 2 and 4 were
younger and leaner at diabetes onset than the other subjects
with KPD (14). All homozygous R133W carriers experienced
remission of insulin dependence within the first 6 months of
initial admission, allowing the maintenance of good blood
glucose control with oral hypoglycemic agents (OHA).
However, the clinical course of patients 3 and 4 was char-
acterized by a new episode of acute insulin deficiency with
development of permanent insulin dependence in ,2 years
after diabetes onset (14). Patients 1 and 3 were still being
treated with OHA 3 years after diabetes onset. Other clinical
and metabolic characteristics at diabetes onset were not differ-
ent between ketosis-prone diabetic carriers of the R133W and
wild-type PAX4 (Table 4).
Patient 5 (R37W) did not experience remission of insulin
dependence and remained insulin dependent from diabetes
onset up to the present time, consistent with a severe b-cell
insulin secretory defect (14).
PAX4 variant carriers show severe alterations
of insulin secretion in vivo
Insulin secretory reserve was assessed in patients 1–5, using a
glucagon-stimulation test as described (9,14,18).
As previously described (14), we observed a relative restor-
ation of insulin secretion following resolution of hyperglyce-
mia in patients 1–4, the homozygous carriers of the R133W
mutation. However, the insulin secretory reserve was more
severely altered than in other ketosis-prone diabetic subjects
with heterozygous R133W or wild-type PAX4 variant
(Fig. 2). In patient 5 (R37W), the glucagon-stimulation test
was consistent with total insulin deficiency [C-peptide
(ng/ml) basal: 0.1; 8 min: 0.6] and insulin secretory reserve
remained low for up to 10 years following admission (data
not shown). It should be noted that although three out of the
four R133W homozygote carriers were from Senegal, the
country of origin did not act as a confounder and we did not
detect any difference in insulin secretion assessed by glucagon-
stimulation, when comparing ketosis-prone diabetic wild-type
carriers from Senegal with those originating from other west
African countries (data not shown).
Thus, both the homozygous R133W and R37W variants are
associated with severe alteration of pancreatic b-cell function
and insulin secretion in vivo.
Separating the common form of type 2 diabetes into pheno-
typic subgroups has led to the identification of MODY, due
to mutations in at least five genes, including glucokinase and
b-cell transcription factors (1,2,4,19) and mitochondrial dia-
betes, a maternally inherited diabetic syndrome, secondary
to a mutation in the mitochondrial genome (20).
KPD is a rare subtype of type 2 diabetes characterized by a
severe b-cell dysfunction and mostly found in populations of
west African ancestry (8–10,12,14,18). Although genetic sus-
ceptibility to KPD is very likely (8–10,12,14,18), it is not
known whether we are facing a polygenic model or a model
with a major gene influence. The observation that targeted dis-
ruption of the PAX4 gene in mice leads to absence of pancrea-
tic b-cells along with the recent finding that PAX4 is a
candidate gene associated with insulin deficient type 2 dia-
betes in Japanese subjects, prompted us to screen this gene
in west Africans with KPD.
Stringent phenotyping has allowed us to identify a popu-
lation of diabetic patients of west African ancestry with well
defined KPD. Using this population, we identified a novel
functional variant in the PAX4 gene (R133W) that is specific
to this population and predisposes to KPD under a recessive
The heterozygous R133W variant is present in 7–10% of
the population of west African descent and is not found in
Caucasians. Similarly, other investigators failed to find this
variant in Japanese (7,15). The high frequency of the hetero-
zygous R133W variant within the west African population
could be due to population history, but could also reflect a
selective advantage conferred to carriers of at least one copy
of the variant, for a disease exposure that is endemic to this
The homozygous R133W variant of PAX4 is present in 4%
of subjects with KPD, is specific to that subtype of diabetes
and dramatically increases the risk of disease. Thus, it could
be a marker of severe insulin deficiency in type 2 diabetic
populations of west African descent. We observe an increased
frequency of the heterozygous R133W carriers in the KPD
group compared with other individuals of the same west
African ancestry, suggesting that the heterozygous state may
also predispose to KPD.
Table 3. R133W/P321H haplotypes
(n ¼ 101) and west African controls (n ¼ 255).
3154Human Molecular Genetics, 2004, Vol. 13, No. 24
The R133W mutation is located between the paired domain
and homeodomain, and leads to functional alterations in vitro.
Studies using a luciferase assay show that R133W leads to a
reduction by half of normal PAX4 function as a transcriptional
repressor. Surprisingly, the R133W variant did not affect
DNA-binding affinity as measured in vitro by EMSA. As
this variant does not lie within the defined repressor domain
of PAX4 (16), the decreased transcriptional repression
observed in the R133W variant may result, not from decreased
DNA-binding affinity for the sites that we tested, but possibly
from altered binding specificity, impaired interaction with
other proteins or from a decreased functional stability in vivo.
Consistent with functional studies, clinical investigation in
ketosis-prone diabetic carriers of the homozygous R133W
variant show a more severe alteration of insulin secretion
than heterozygous and wild-type ketosis-prone diabetic car-
riers. Together, the genetic study along with the in vitro and
in vivo functional characterization, suggest that the homozy-
gous R133W variant predisposes to KPD.
We have also identified a R37W variant in one patient.
R37W affects the DNA-binding domain of the molecule and
in vitro shows a 50% reduction of PAX4-binding activity to
target gene promoters resulting in a similar decrease of
normal PAX4 transcriptional repression activity. This more
severe biochemical phenotype of R37W compared with
R133W may help explain why it is clinically apparent even
in its heterozygous form. However, because the R37W
variant was found in only one case, we cannot ascertain its
role in diabetes without family studies, and these have not
A variant in PAX4 (R121W) has been described previously
in the Japanese type 2 diabetic population. Interestingly, the
Japanese R121W diabetic carriers and west African homozy-
gous R133W diabetic carriers share some clinical similarities:
half the type 2 diabetic carriers of the R121W variant showed
initial insulin dependence (14). In addition, the homozygous
R121W carrier had early-onset diabetes (29 years old) and
early insulin dependence (7), and half of the west African
homozygous R133W ketosis-prone diabetic carriers were
under 25 years of age, at diabetes onset. In vitro functional
studies of the R121W Japanese variant reveal a more severe
alteration of transcriptional repression of target genes and
DNA-binding activity compared with the west African
R133W variant. This is consistent with the observation that
the 121W allele is absent from the background Japanese popu-
lation (strong effect in heterozygotes), whereas the 133W
Figure 1. (A) Structure of the PAX4 molecule and position of the R37W and R133W variants. (B) Effect of the PAX4 mutations on gene transcription activity in
a-TC1.6 cells. The wild-type and mutant hPAX4 (WT, R133W, R37W) were co-transfected with an insulin promoter reporter construct (2410Rins-pfoxluc) in
a-TC1.6 cells. The amount of reporter plasmid used was 2 mg/lane (lanes 1–5). The amount of co-transfected hPAX4 constructs (wt and variants) was 50 ng/lane
(lanes 3–5). Data are expressed as mean+ SE.?P , 0.01 comparing R133W or R37W to WT. (C) Effect of PAX4 mutations on PAX4 binding to target genes.
An EMSA using control (lane 1), wild-type PAX4 (lane 2), R37W variant (lane 3) and R133W variant (lane 4) is shown. The arrow indicates binding of PAX4
proteins to the probe. The oligonucleotide probes were the human PAX4 promoter (left) and the rat insulin promoter (right). Identical results were obtained in at
least two independent experiments.
Human Molecular Genetics, 2004, Vol. 13, No. 24 3155
Table 4. Clinical characteristics of the R37W and R133W variant carriers
Homozygous R133W (n ¼ 4)
Patient 1Patient 2
Heterozygous R133W (n ¼ 27)
Heterozygous R37W (n ¼ 1)
Wild-type PAX4 (n ¼ 70)
Patient 3Patient 4Mean+ SDa
Family history of diabetes
ICA, GAD 65 ab
Age at onset (years)
BMI (kg/m2) at admission
BMI (kg/m2) before symptoms
Weight loss (kg)
Blood glucose (mM)
HbA1c (%) at admission
C-peptide at admission (ng/ml)
Glucagon at admission (pg/ml)c
Remission of insulin dependence
Duration of initial
Insulin therapy (weeks)
Treatment after remission of
HbA1c (%) at remission
Relapse in insulin dependence
aMean+ SD for R133W homozygous carriers.
bDifference between R133W homozygous carriers and wild-type carriers (t-test). Normal HbA1c is ,6%. OHA, oral hypoglycemic agent.
cSerum glucagon values in non-diabetic individuals: 40–130 pg/ml.
Human Molecular Genetics, 2004, Vol. 13, No. 24
allele is present in 7–10% of the population of west African
ancestry (weak effect in heterozygotes). Taken together,
these studies suggest that PAX4 is a candidate gene for
severe insulin deficiency in non-white populations.
Evidence from mice with targeted disruption of the PAX4
gene demonstrates a role for PAX4 in b-cell formation
during fetal development (21). In these mice, loss of PAX4
decreases the expression of b-cell specific genes and results
in a dramatic decrease in b-cell mass at birth (22). Thus,
homozygous R133W carriers may have fewer b-cells at
birth, or may have difficulty replacing b-cell mass as they
get older. Alternatively, inadequate PAX4 function could
lead to the formation of dysfunctional b-cells.
This study presents two limitations: One is the lack of
access to family members at inclusion to demonstrate the seg-
regation of the mutations with KPD. The other is the lack of
power of our sample size. As west Africans are the genetic
founders of African-Americans, a large study is needed to
explore the role of PAX4 variants in African-Americans
with KPD at the level of the US population.
In summary, we describe a novel functional variant of
PAX4 (R133W) that is specific to west Africans. Homozygous
carriers of the PAX4 R133W variant have an increased risk of
KPD. These data provide the first evidence that ethnic-specific
gene variants may contribute to the predisposition to this par-
ticular form of diabetes and suggest that KPD, like MODY, is
a rare phenotypically defined but genetically heterogeneous
form of type 2 diabetes.
MATERIALS AND METHODS
Patient and control populations
The diabetic populations used in this study have been
described previously (14,23). Briefly, KPD was defined as
new onset diabetes, with the presence of strong ketosis
(urine ketones .80 mg/dl) or DKA, without precipitating
illness (infection, stress), and in the absence of auto-antibodies
(ab) to islet cells (ICA) and to glutamic acid decarboxylase
(GAD) 65. Ketosis-prone diabetic patients did not show the
classical HLA-DR haplotypes encountered in type 1 diabetes
(18). Type 1 diabetic subjects were defined by the presence
of ICAs and GAD 65 ab. Type 2 diabetic subjects were
selected for diabetes treated by diet or OHA and were resistant
to ketosis despite the presence of precipitating illness (14,23).
The control populations come from either the Lariboisiere/
St Louis Medical Center, outpatient clinic, Paris, France (west
Africans) or the Bermondsey and Lansdowne Medical
Mission, the Clapham Manor Health Centre, and from
two churches in south-east London, UK (west Africans,
without family history of diabetes. The diabetic and control
populations were comparable with regard to their west
African countries of origin. Other controls (Caucasians)
come from the human variation collection of the NIGMS repo-
sitory (Coriell Institute for Medical Research, Camden, NJ).
Informed consent was obtained from all subjects. Genomic
DNA was extracted from peripheral lymphocytes for all sub-
jects as previously described (23). This protocol was approved
by all the local ethics committees.
Pancreatic insulin reserve was assessed at least 48 h after res-
olution of ketosis and normalization of blood glucose by
measuring the C-peptide level before and 8 min following
the i.v. injection of 1 mg glucagon as previously described
(14). Patients were studied 12 h after an overnight fast and
before taking the morning insulin injection.
Blood was drawn during the initial admission for diabetic
ketoacidosis in tubes containing aprotinin as protease inhibi-
tor. Serum glucagon level was assessed using a glucagon
RIA kit (Linco Research, St Charles, MO, USA) and a
glucagon-specific antibody with a sensitivity of 20 pg/ml.
A standard curve was generated with purified recombinant
glucagon over a range of 20–400 pg/ml.
PAX4 mutation screening
The coding sequence of the human PAX4 gene (nine exons)
was amplified in three different PCR reactions: Exon 1, 2, 3:
PAX4-1F (AGG TGG TGT GTG GAT ACC TC) and
PAX4-3R (GAT TTG GCT GTG ATT AGC CC); Exon 4,
5, 6, PAX4-4F (CTG ACC AGA GGA ATC ACC ATC)
and PAX4-6R (GAT GAC TGA GCG GGC AGA TG);
Exon 7, 8, 9: PAX4-7F (AGT GGC TGA CTT TCC TAG
AAC) and PAX4-9R (TGG GCA GGA TGG TAT TAG
ATC TTC TCT ATG). Sequencing reactions were performed
with the BigDye terminator kit (Applied Biosystems, Foster
City, CA, USA) under the standard manufacturer’s conditions.
Sequencing was performed on an ABI PRISMw3700
automated DNA sequencer (Applied Biosystems). Sequences
wereanalyzed using Sequence
Figure 2. b-Cell insulin secretory reserve was assessed in a cohort of
ketosis-prone diabetic subjects carrying the wild-type PAX4 (R/R, n ¼ 18),
the heterozygote (R/W, n ¼ 11) and the homozygous mutation (W/W,
n ¼ 4), by measuring basal C-peptide (0 min) and C-peptide response follow-
ing IV glucagon injection (8 min) after correction of hyperglycemia.
?P , 0.05, R/R versus W/W.
Human Molecular Genetics, 2004, Vol. 13, No. 243157
PAX4 variants genotyping
The C397T [R133W] mutation creates a restriction site for the
enzyme Hsp92 II that we used to create a restriction length
fragment polymorphism (RFLP) assay. We amplified PAX4
exon 3 from genomic DNA by PCR, using primers
PAX4-3F (AGCCCTGAGTCTGAGCACCA) and PAX4-
3R2 (GGAGAGAATGAGACTCCCT). Restriction digest of
the 260 bp PCR product by the enzyme Hsp92 II (Promega,
Wisconsin, MA, USA) was performed as recommended by
the manufacturer and lead to 110 and 150 bp products in
presence of the mutant allele. Homozygous mutations were
confirmed by direct sequencing.
Other PAX4 variants were genotyped by TaqMan allelic
discrimination assay system after designing flanking primers
and fluorogenic probes, both of which were specific to target
Cloning of wild-type and mutant PAX4
A PCDNA3.1 plasmid expressing human PAX4 (hPAX4) was
used to clone the mutations C109T (Arg37Trp) and C397T
(Arg133Trp) by site-directed mutagenesis. Both mutations are
flanked by the unique restriction sites AflII and BsteII,
passing these respective restriction sites and used these
primers to amplify large PCR products that were cloned back
into the wild-type PCDNA3.1 backbone.
Electromobility shift assays
For in vitro expression, the mutant and wild-type hPAX4 were
cloned into the plasmid pSP72 downstream of the SP6 promo-
ter. The entire PAX4 sequence of all constructs was verified
by direct sequencing. The wild-type and mutant PAX4 pro-
teins were produced in vitro using the TNT Quick Coupled
Lysate Systemw(Promega) and 1 ml of the 50 ml total reaction
volume was used per binding mix.
Single stranded oligonucleotides were 50-end labelled with
[g-32P]ATP using T4 polynucleotide kinase (16). An excess
of complementary strand was then annealed to form a duplex
strand that was column purified. For EMSA buffers and
electrophoresis conditions, we used 500 ng of poly(dI–dC):
poly(dI–dC) per 10 ml binding mix as described previously
(16). Oligonucleotides used were as follows (coding strand
shown from each double-stranded pair): Rat insulin I C2
element, bp 2328 to 2304 50-ctgggaaatgaggtggaaaatgctc-30;
Human PAX4 promoter, bp 24164 to 24116 (P4 4.2) 50-
cccaattgtcaaaggtggaataatttgatcaacaaaataatgtattg-30. EMSA results
are representative of at least two independent experiments.
Cell culture and transient transfections
a-TC1.6 cells were grown in Dulbecco’s modified Eagle’s
medium (DMEM) supplemented with 2.5% fetal bovine
serum and 15% horse serum. Twenty-four hours prior to trans-
fection,cells were split into6-wellplates and1millioncells per
well were used for transfection. The reporter construct used
(2410Rins-pfoxluc) was created using the 2410 bp of the rat
insulin promoter cloned upstream of luciferase in the
pFoxluc1 backbone reporter vector. An aliquot of 2 mg of the
reporter construct was used per well, 50 ng of any co-
fast lipidagent(Promega)was usedforalltransfections accord-
ing to the manufacturer’s instructions. Cells were harvested
48 h after transfection and luciferase assays performed as pre-
viously described (24). Luciferase activity was corrected for
protein concentration; at least three independent sets of trans-
fections were performed. Data are expressed as mean + SE.
For association studies of the R133W variant, each diabetic
and control group was tested for deviations from Hardy–
Weinberg Equilibrium using x2tests. The minor allele
frequency was estimated for each group and the 95% CI was
calculated on the basis of the binomial distribution. In order
to test whether there was a significant difference in frequency
of homozygous R133W individuals in KPD compared with
control individuals, the ketosis-prone diabetic subjects were
compared with the west African control groups and a Fisher-
exact test was performed. The OR was also calculated using
a Sheehe correction (25) owing to the presence of a cell
with 0 observation in the non-ketosis-prone diabetic groups.
Owing to ambiguity of heterozygous individuals at both
loci, haplotypes were constructed for ketosis-prone diabetic
subjects and west African controls using the SNPHAP
txt) program which implements the Expectation Maximization
(EM) algorithm. Using the estimated haplotype frequencies
under linkage equilibrium and disequilibrium, D0was esti-
mated for both the cases and controls (26). A likelihood
ratio test was performed to determine if there was a difference
in the haplotype frequencies between cases and controls.
Briefly, the log-likelihood of the cases and controls estimated
separately was compared to the log-likelihood of the entire
dataset [2(ln(L, cases) þ ln(L, controls)) 2 ln(L, cases þ
controls together)]. Differences in haplotype frequencies
(R133W/P321H) were evaluated between the ketosis-prone
diabetic cases and the other groups by calculating ORs and
their 95% CI using the Sheehe correction. The x2test was
used to evaluate if the differences in haplotype frequencies
between the ketosis-prone diabetic cases and controls were
For in vitro and in vivo studies, each variable was analyzed
using the unpaired Student’s test. For all analyses, a P-value of
,0.05 was considered significant. Results are given as
We gratefully acknowledge F. Andreelli, M. Assayag,
F. Bosquet, C. Carette, B. Charbonnel, G. Charpentier,
A. Colombel, N. D’Escrivain, C. Deybach, F. Duron,
A. Grimaldi, P.J. Guillausseau, H. Hanaire-Broutin, J. Samuel-
Lajeunesse, N. Vigier, M.L. Virally, A. Warnet and the many
3158Human Molecular Genetics, 2004, Vol. 13, No. 24
We thank Jean-Marie Villette for DNA extraction. F.M.-J. Download full-text
was supported by the Assistance Publique-Ho ˆpitaux de
Paris, an unrestricted grant from Bristol-Myers Squibb and
Novo-Nordisk, and by the Division of Diabetes, Endocrinology
and Metabolism at Baylor College of Medicine. C.V. was
supported by NIH RO1 DK60540 and an American Diabetes
Association Career Development Award.
1. Yamagata, K., Furuta, H., Oda, N., Kaisaki, P.J., Menzel, S., Cox, N.J.,
Fajans, S.S., Signorini, S., Stoffel, M. and Bell, G.I. (1996) Mutations in
the hepatocyte nuclear factor-4alpha gene in maturity-onset diabetes of
the young (MODY1). Nature, 384, 458–460.
2. Yamagata, K., Oda, N., Kaisaki, P.J., Menzel, S., Furuta, H.,
Vaxillaire, M., Southam, L., Cox, R.D., Lathrop, G.M., Boriraj, V.V. et al.
(1996) Mutations in the hepatocyte nuclear factor-1alpha gene in
maturity-onset diabetes of the young (MODY3). Nature, 384, 455–458.
3. Stoffers, D.A., Ferrer, J., Clarke, W.L. and Habener, J.F. (1997)
Early-onset type-II diabetes mellitus (MODY4) linked to IPF1.
Nat. Genet., 17, 138–139.
4. Horikawa, Y., Iwasaki, N., Hara, M., Furuta, H., Hinokio, Y.,
Cockburn, B.N., Lindner, T., Yamagata, K., Ogata, M., Tomonaga, O.
et al. (1997) Mutation in hepatocyte nuclear factor-1 beta gene (TCF2)
associated with MODY. Nat. Genet., 17, 384–385.
5. Malecki, M.T., Jhala, U.S., Antonellis, A., Fields, L., Doria, A., Orban, T.,
Saad, M., Warram, J.H., Montminy, M. and Krolewski, A.S. (1999)
Mutations in NEUROD1 are associated with the development of type 2
diabetes mellitus. Nat. Genet., 23, 323–328.
6. Sosa-Pineda, B., Chowdhury, K., Torres, M., Oliver, G. and Gruss, P.
(1997) The Pax4 gene is essential for differentiation of insulin-producing
beta cells in the mammalian pancreas. Nature, 386, 399–402.
7. Shimajiri, Y., Sanke, T., Furuta, H., Hanabusa, T., Nakagawa, T.,
Fujitani, Y., Kajimoto, Y., Takasu, N. and Nanjo, K. (2001) A missense
mutation of Pax4 gene (R121W) is associated with type 2 diabetes in
Japanese. Diabetes, 50, 2864–2869.
8. Banerji, M.A., Chaiken, R.L., Huey, H., Tuomi, T., Norin, A.J.,
Mackay, I.R., Rowley, M.J., Zimmet, P.Z. and Lebovitz, H.E. (1994)
GAD antibody negative NIDDM in adult black subjects with diabetic
ketoacidosis and increased frequency of human leukocyte antigen DR3
and DR4. Flatbush diabetes. Diabetes, 43, 741–745.
9. Umpierrez, G.E., Casals, M.M., Gebhart, S.P., Mixon, P.S., Clark, W.S.
and Phillips, L.S. (1995) Diabetic ketoacidosis in obese African-
Americans. Diabetes, 44, 790–795.
10. Pinero-Pilona, A., Litonjua, P., Aviles-Santa, L. and Raskin, P. (2001)
Idiopathic type 1 diabetes in Dallas, Texas: a 5-year experience. Diabet.
Care, 24, 1014–1018.
11. Sobngwi, E., Mauvais-Jarvis, F., Vexiau, P., Mbanya, J.C. and
Gautier, J.F. (2002) Diabetes in Africans. Part 2: Ketosis-prone atypical
diabetes mellitus. Diabet. Metab., 28, 5–12.
12. Maldonado, M., Hampe, C.S., Gaur, L.K., D’Amico, S., Iyer, D.,
Hammerle, L.P., Bolgiano, D., Rodriguez, L., Rajan, A., Lernmark, A.
et al. (2003) Ketosis-prone diabetes: dissection of a heterogeneous
syndrome using an immunogenetic and beta-cell functional classification,
prospective analysis, and clinical outcomes. J. Clin. Endocrinol. Metab.,
13. Kitabchi, A.E. (2003) Ketosis-prone diabetes—a new subgroup of patients
with atypical type 1 and type 2 diabetes? J. Clin. Endocrinol. Metab., 88,
14. Mauvais-Jarvis, F., Sobngwi, E., Porcher, R., Riveline, J.P.,
Kevorkian, J.P., Vaisse, C., Guillausseau, P.J., Charpentier, G., Vexiau, P.
and Gautier, J.F. (2004) Ketosis-prone type 2 diabetes in patients of
sub-Saharan African origin: clinical pathophysiology and natural history
of beta-cell dysfunction and insulin resistance. Diabetes, 53, 645–653.
15. Dupont, S., Vionnet, N., Chevre, J.C., Gallina, S., Dina, C., Seino, Y.,
Yamada, Y. and Froguel, P. (1999) No evidence of linkage or diabetes-
associated mutations in the transcription factors BETA2/NEUROD1 and
PAX4 in Type II diabetes in France. Diabetologia, 42, 480–484.
16. Smith, S.B., Ee, H.C., Conners, J.R. and German, M.S. (1999) Paired-
homeodomain transcription factor PAX4 acts as a transcriptional repressor
in early pancreatic development. Mol. Cell. Biol., 19, 8272–8280.
17. Smith, S.B., Watada, H., Scheel, D.W., Mrejen, C. and German, M.S.
(2000) Autoregulation and maturity onset diabetes of the young
transcription factors control the human PAX4 promoter. J. Biol. Chem.,
18. Sobngwi, E., Vexiau, P., Levy, V., Lepage, V., Mauvais-Jarvis, F.,
Leblanc, H., Mbanya, J.C. and Gautier, J.F. (2002) Metabolic and
immunogenetic prediction of long-term insulin remission in African
patients with atypical diabetes. Diabet. Med., 19, 832–835.
19. Froguel, P., Zouali, H., Vionnet, N., Velho, G., Vaxillaire, M., Sun, F.,
Lesage, S., Stoffel, M., Takeda, J., Passa, P. et al. (1993) Familial
hyperglycemia due to mutations in glucokinase. Definition of a subtype
of diabetes mellitus. N. Engl. J. Med., 328, 697–702.
20. Kadowaki, T., Kadowaki, H., Mori, Y., Tobe, K., Sakuta, R., Suzuki, Y.,
Tanabe, Y., Sakura, H., Awata, T., Goto, Y. et al. (1994) A subtype of
diabetes mellitus associated with a mutation of mitochondrial DNA.
N. Engl. J. Med., 330, 962–968.
21. Prado, C.L., Pugh-Bernard, A.E., Elghazi, L., Sosa-Pineda, B. and
Sussel, L. (2004) Ghrelin cells replace insulin-producing beta cells in
two mouse models of pancreas development. Proc. Natl Acad. Sci.
USA, 101, 2924–2929.
22. Wang, J., Elghazi, L., Parker, S.E., Kizilocak, H., Asano, M., Sussel, L.
and Sosa-Pineda, B. (2004) The concerted activities of Pax4 and Nkx2.2
are essential to initiate pancreatic beta-cell differentiation. Dev. Biol., 266,
23. Mauvais-Jarvis, F., Boudou, P., Sobngwi, E., Riveline, J.P., Kevorkian,
J.P., Villette, J.M., Porcher, R., Vexiau, P. and Gautier, J.F. (2003) The
polymorphism Gly574Ser in the transcription factor HNF-1alpha is not a
marker of adult-onset ketosis-prone atypical diabetes in Afro-Caribbean
patients. Diabetologia, 46, 728–729.
24. German, M.S., Wang, J., Chadwick, R.B. and Rutter, W.J. (1992)
Synergistic activation of the insulin gene by a LIM-homeo domain
protein and a basic helix–loop–helix protein: building a functional insulin
minienhancer complex. Genes Dev., 6, 2165–2176.
25. Sheehe, P.R. (1966) Combination of log relative risk in retrospective
studies of disease. Am. J. Public Health Nations Health, 56,
26. Lewontin, R.C. (1964) The interaction of selection and linkage: I. General
considerations. Genetics, 49, 49–67.
Human Molecular Genetics, 2004, Vol. 13, No. 243159