Human CD38 is an authentic NAD(P)+ glycohydrolase

Abstract

The leucoyte surface antigen CD38 has been shown to be an ecto-enzyme with multiple catalytic activities. It is principally a NAD+ glycohydrolase that transforms NAD+ into ADP-ribose and nicotinamide. CD38 is also able to produce small amounts of cyclic ADP-ribose (ADP-ribosyl cyclase activity) and to hydrolyse this cyclic metabolite into ADP-ribose (cyclic ADP-ribose hydrolase activity). To classify CD38 among the enzymes that transfer the ADP-ribosyl moiety of NAD+ to a variety of acceptors, we have investigated its substrate specificity and some characteristics of its kinetic and molecular mechanisms. We find that CD38-catalysed cleavage of the nicotinamide-ribose bond results in the formation of an E.ADP-ribosyl intermediary complex, which is common to all reaction pathways; this intermediate reacts (1) with acceptors such as water (hydrolysis), methanol (methanolysis) or pyridine (transglycosidation), and (2) intramolecularly, yielding cyclic ADP-ribose with a low efficiency. This reaction scheme is also followed when using nicotinamide guanine dinucleotide as an alternative substrate; in this case, however, the cyclization process is highly favoured. The results obtained here are not compatible with the prevailing model for the mode of action of CD38, according to which this enzyme produces first cyclic ADP-ribose which is then immediately hydrolysed into ADP-ribose (i.e. sequential ADP-ribosyl cyclase and cyclic ADP-ribose hydrolase activities). We show instead that the cyclic metabolite was a reaction product of CD38 rather than an obligatory reaction intermediate during the glycohydrolase activity. Altogether our results lead to the conclusion that CD38 is an authentic 'classical' NAD(P)+ glycohydrolase (EC 3.2.2.6).

Full text

Biochem. J. (1998) 330, 1383–1390 (Printed in Great Britain) 1383
Human CD38 is an authentic NAD(P)
+
glycohydrolase
Vale
!
rie BERTHELIER*, Jean-Michel TIXIER*, He
!
le
'
ne MULLER-STEFFNER†, Francis SCHUBER† and Philippe DETERRE*
1
*Laboratoire d’Immunologie Cellulaire, Unite
!
Associe
!
e 625 du Centre National de la Recherche Scientifique, Groupe Hospitalier Pitie
!
-Salpe
#
tie
'
re, 83 boulevard de l’Ho
#
pital,
75013 Paris, France, and †Laboratoire de Chimie Bioorganique, Unite
!
Associe
!
e 1386 du Centre National de la Recherche Scientifique, Faculte
!
de Pharmacie, 74 route
du Rhin, 67400 Illkirch, France
The leucoyte surface antigen CD38 has been shown to be an
ecto-enzyme with multiple catalytic activities. It is principally a
NAD
+
glycohydrolase that transforms NAD
+
into ADP-ribose
and nicotinamide. CD38 is also able to produce small amounts
of cyclic ADP-ribose (ADP-ribosyl cyclase activity) and to
hydrolyse this cyclic metabolite into ADP-ribose (cyclic ADP-
ribose hydrolase activity). To classify CD38 among the enzymes
that transfer the ADP-ribosyl moiety of NAD
+
to a variety of
acceptors, we have investigated its substrate specificity and some
characteristics of its kinetic and molecular mechanisms. We find
that CD38-catalysed cleavage of the nicotinamide-ribose bond
results in the formation of an E[ADP-ribosyl intermediary
complex, which is common to all reaction pathways ; this
intermediate reacts (1) with acceptors such as water (hydrolysis),
methanol (methanolysis) or pyridine (transglycosidation), and
INTRODUCTION
The human cell-surface antigen CD38 is a 46 kDa type II
glycoprotein with a single transmembrane domain [1]. Its ex-
pression is widely used as a phenotypic marker of differentiation
or activation of human T and B lymphocytes [2–4]. The high
sequence similarity between ADP-ribosyl cyclase from Aplysia
californica and CD38 [5] led to the finding that this antigen, for
which no biological activity is yet known, is an ectoenzyme that
catalyses the cleavage of the nicotinamide-ribose bond in NAD
+
(reviewed in [2–4]). In contrast with the invertebrate enzyme,
which transforms NAD
+
exclusively into cyclic ADP-ribose
(cADPR), human and murine CD38 were reported to be endowed
with at least three different catalytic activities: NAD
+
glyco-
hydrolase (NADase) and ADP-ribosyl cyclase activities that
convert NAD
+
into nicotinamide and respectively ADP-ribose
and cADPR, and a cADPR hydrolase activity producing ADP-
ribose from cADPR [6–10]. CD38 is prevalently an NADase and
only very little cADPR is formed during the transformation of
NAD
+
, i.e. less than 1% of reaction products [6–9,11]. The low
yield of the cyclic metabolite, which is thought to be an
endogenous regulator of the Ca
#
+
-induced Ca
#
+
-release process
mediated by the ryanodine receptor [12–14], was generally
attributed to the ‘ multifunctionality’ of CD38 [3,6,10,15,16].
Thus it was proposed that the measurable NADase activity of
CD38 resulted from the addition of its ADP-ribosyl cyclase and
cADPR hydrolase activities, with cADPR as a reaction in-
termediate that is quickly turned over [6,10,15,16]. This as-
Abbreviations used: εNAD
+
, nicotinamide 1,N
6
-adenine dinucleotide; cADPR, cyclic ADP-ribose (ADP-cyclo[N
1
,C-1§]-ribose); NGD
+
, nicotinamide
guanine dinucleotide ; cGDPR, cyclic GDP-ribose (GDP-cyclo[N
7
,C-1§]-ribose) ; INH, isonicotinic acid hydrazide; hy
4
PyAD
+
, 4-hydrazinocarbonylpyridine
adenine dinucleotide; araF-NAD
+
, nicotinamide 2«-deoxy-2«-fluoroarabinoside adenine dinucleotide; NADase, NAD
+
glycohydrolase.
1
To whom correspondence should be addressed.
(2) intramolecularly, yielding cyclic ADP-ribose with a low
efficiency. This reaction scheme is also followed when using
nicotinamide guanine dinucleotide as an alternative substrate; in
this case, however, the cyclization process is highly favoured.
The results obtained here are not compatible with the prevailing
model for the mode of action of CD38, according to which this
enzyme produces first cyclic ADP-ribose which is then im-
mediately hydrolysed into ADP-ribose (i.e. sequential ADP-
ribosyl cyclase and cyclic ADP-ribose hydrolase activities). We
show instead that the cyclic metabolite was a reaction product of
CD38 rather than an obligatory reaction intermediate during the
glycohydrolase activity. Altogether our results lead to the con-
clusion that CD38 is an authentic ‘ classical’ NAD(P)
+
glyco-
hydrolase (EC 3\2\2\6).
sumption strongly questioned the real catalytic activity of the
‘classical’ NADases (EC 3\2\2\5 and 3\2\2\6) which, in
mammals, constitute a heterogeneous family of enzymes (in
terms of molecular mass and catalytic properties) [17]. The
question arose whether the NADases were also multifunctional
enzymes whose ADP-ribosyl cyclase and cADPR hydrolase
activities had been overlooked. Indeed, Kim et al. [18] found a
canine spleen NADase that possessed also cADPR-producing
and -degrading activities. Recent studies by Schuber’s group
established that the well-known calf spleen ecto-NADase was
also able to catalyse the transformation of NAD
+
into small
amounts of cADPR and to hydrolyse cADPR ; however, their
studies on the cyclization mechanism, which excluded cADPR as
a kinetically competent reaction intermediate in the transform-
ation of NAD
+
into ADP-ribose [19], did not support the
mechanism that was suggested for CD38 [20]. The present
situation is therefore rather confusing and one can therefore
wonder whether CD38 and the classical ecto-NADases are the
same enzymes or whether they constitute subclasses of a super-
family of enzymes that might differ in some aspects of their
molecular mechanism. Related to this question is the observation
that mammalian NADases have a much wider tissue and cellular
distribution [17] than originally thought for CD38, although this
latter point is rapidly evolving [3,9,21–24]. In the absence of
structural data for NADases, it was therefore of importance to
characterize more accurately the catalytic properties of CD38 in
accordance with criteria used to classify the different NADases
[17]. From our studies it seems likely that human CD38 is an

1384 V. Berthelier and others
authentic ecto-NAD(P)
+
glycohydrolase, belonging to the EC
3\2\2\6 class of nucleosidases that hydrolytically cleave both
NAD
+
and NADP
+
.
EXPERIMENTAL
Chemicals
All chemicals used were from Sigma (Saint Quentin Fallavier,
France).
Source of CD38
Intact human B Daudi cells were used as a source of CD38. This
cell line expresses CD38 highly and lacks other ectoenzymes,
such as nucleotide pyrophosphatase, that also metabolize NAD
+
.
Most studies reported here were performed on CD38 purified by
immunoprecipitation (see below), preliminary experiments
having shown that CD38 ligation with antibodies did not alter its
enzymic activity. To check the purity of this preparation, an
immunoprecipitate prepared from biotinylated Daudi cells was
subjected to SDS}PAGE, blotted on nitrocellulose and labelled
with avidin: one major band at 45 kDa was obtained, in
agreement with similar preparations reported previously [25]. A
second source of CD38 was a recombinant fusion protein between
the extracellular portion of human CD38 and mouse CD8α [26]
kindly provided by Dr. Banchereau and Dr Brie
'
re (Schering
Plough, Dardilly, France). This preparation consists of a super-
natant of transfected COS cells and was estimated as 95 % pure
by SDS}PAGE stained with Coomassie Blue. The specific
activities, at pH 7±4 and 37 °C, were: 5 m-unit per 10
'
Daudi
cells, 0±3 m-unit for an immunoprecipitate prepared from 10
'
Daudi cells and 10 m-unit}mg of protein for the recombinant
CD38. One unit of CD38 NADase activity is defined as the
amount able to hydrolyse 1 µmol of NAD
+
}min.
Cell culture and CD38 immunopurification
Daudi cells were grown in RPMI 1640 medium supplemented
with 10% (v}v) fetal calf serum, 2 mM -glutamine, 1 mM
sodium pyruvate, 50 i.u.}ml penicillin and 50 µg}ml strepto-
mycin, at a density of (0±5–1)¬10
'
cells}ml. For CD38 immuno-
preparation, 10
)
cells were washed twice in PBS and incubated
for 30 min at 4 °C in 1 ml of lysis buffer [20 mM Tris (pH 7±5)}1
mM EDTA}140 mM NaCl}1 % (v}v) Nonidet P40 0±5 %
aprotinin}1 mM PMSF]. Cell debris was removed by centri-
fugation at 13000 g for 15 min and the supernatant was pre-
cleared by incubation for 40 min at 4 °C with 10 mg of Protein A
immobilized on Q Sepharose CL-4B. The supernatant was then
incubated with 1 % (v}v) CD38-specific monoclonal antibodies
BB51 (ascites fluid kindly provided by Dr. L. Boumsell) [27] at
4 °C for 2 h. Protein A–Sepharose (10 mg) was then added to the
reaction mixture, followed by further incubation at 4 °C for 1 h.
Immune complexes were washed three times with lysis buffer and
once with buffer A (see below), before being used in enzymic
assays.
Enzymic assays and HPLC analysis of reactions products
Reactions were performed in buffer A [50 mM Hepes (pH
7±4)}150 mM NaCl}1 mM CaCl
#
}0±5 mM MgCl
#
}5 mM KCl}1
mM Na
#
HPO
%
]at37°C. At given times 50 µl aliquots were taken
and centrifuged briefly (5000 g for 10 s). For reactions with
recombinant CD38 protein, aliquots were frozen in liquid N
#
.
Samples were then diluted by addition of 500 µl of the starting
HPLC buffer (see below) and filtered on 0±2 µm (pore size)
cellulose filters before injection. For methanolysis and trans-
glycosidation experiments, chromatography was performed on
a 300 mm¬3±9mmµBondapak C
")
column (Waters), at a flow
rate of 1 ml}min, with a Beckman system equipped with a Gold-
166 spectrophotometric detector set at 260 nm. Compounds were
eluted isocratically with 10 mM ammonium phosphate buffer,
pH 5±6, containing a given percentage of acetonitrile [i.e. 1±5 %
(v}v) for the elution of products obtained by hydrolysis and
methanolysis of NAD
+
, nicotinamide 1,N
'
-adenine dinucleotide
(ε-NAD
+
), cADPR and nicotinamide guanine dinucleotide
(NGD
+
); 2% (v}v) for the elution of the products obtained by
transglycosidation with isonicotinic acid hydrazide (INH); and
2±5% (v}v) for the elution of the products obtained by trans-
glycosidation with 3-aminopyridine and 3-acetylpyridine].
To assay the hydrolytic capacity of the CD38 preparations, at
least four aliquots of the reaction medium were sequentially
taken, analysed by HPLC and the slopes of the substrate
disappearance or product formation progression curves were
calculated (means³S.D.). Reaction products were identified by
coelution with authentic samples. The mixture of α- and
β-methyl ADP-ribose was prepared by chemical solvolysis
of NAD
+
; β-methyl ADP-ribose was obtained by the treatment of
NAD
+
with purified calf spleen NADase in the presence of
methanol, as reported previously [28]. Kinetic parameters (K
m
, V )
were obtained by analysis, with a non-linear regression program,
of the initial rates plotted against substrate concentrations, with
using at least five data points.
Fluorometric assay of NADase and GDP-ribose cyclase activities
NADase activity was also assayed fluorometrically with ε-NAD
+
as substrate as previously described [29]. Briefly, ε-NAD
+
was
added to the assay medium (buffer A, 2 ml final volume)
containing CD38 (cells, immunoprecipitate or recombinant) at
37 °C in a thermostatically controlled fluorimeter (Perkin-Elmer,
Bois d’Arcy, France) cuvette. The fluorescence emission at 410
nm (excitation at 300 nm) was then followed. To calculate the
initial rates, the maximal fluorescence was obtained by the
addition of 0±1 unit of Crotalus atrox venom pyrophosphatase
(1 unit is the amount able to hydrolyse 1 µmol of NAD
+
}min).
The GDP-ribosyl cyclase activity of CD38 was assayed similarly,
by estimating the appearance of the highly fluorescent cyclic
GDP-ribose (emission at 410 nm, excitation at 300 nm) [30,31].
Slow-binding inhibition of CD38 by nicotinamide 2«-deoxy-2«-
fluoroarabinoside adenine dinucleotide (araF-NAD
+
)
Hydrolysis of ε-NAD
+
was followed by the continuous fluoro-
metric assay, as described above, in the presence of various
concentrations of araF-NAD
+
. About 15–25 points obtained
from the progress curve were taken at times corresponding to
equal increments of fluorescence (F ) and were analysed, as
described previously [32], by fitting to the equation:
F ¯ 
s
t(
!
®
s
)(1®exp(®kt))}kF
!
(1)
Use of a non-linear regression program yielded the different
parameters, i.e. 
!
initial rate; 
s
, steady-state rate ; k, the apparent
first-order constant for reaching the steady-state enzyme–
inhibitor complex; and F
!
, the initial fluorescence. The kinetic
parameters k
off
, k
on
and K
i
(¯ k
off
}k
on
) were calculated from the
plot of k against inhibitor concentrations according to the
equation:
k ¯ k
off
k
on
[I]}(1[S]}K
m
) (2)

1385Human CD38 is an authentic NAD(P)
+
glycohydrolase
Protein assay
Protein concentration was calculated with the Bio-Rad Bradford
protein assay (Ivry sur Seine, France), with BSA as standard.
RESULTS
For the present studies we used immunopurified CD38, orig-
inating from Daudi cells, and the recombinant form expressed in
mammalian cells. Both human CD38 sources had similar catalytic
properties, which were compared with features that are charac-
teristic of ‘classical’ NADases [16].
Substrate specificity
The kinetic parameters of NAD
+
hydrolysis catalysed by CD38
were first determined, followed by the exploration of the substrate
specificity of the enzyme by using β-NAD
+
and analogues (Table
1). The K
m
found with native CD38, i.e. 46 µM, is somewhat
higher than that (15 µM) obtained with a non-glycosylated
recombinant form of CD38 [33]. The enzyme also hydrolytically
cleaved NADP
+
and ε-NAD
+
, a fluorescent analogue of NAD
+
,
with relative V
max
similar to those found previously [11] ; the K
m
for ε-NAD
+
was, however, somewhat lower than that for β-
NAD
+
. In contrast, 3-acetylpyridine adenine dinucleotide, with a
K
m
in the upper µM range, was a poorer substrate for CD38.
Nevertheless under saturating conditions this pyridinium ana-
logue of NAD
+
was hydrolysed almost as fast as NAD
+
. These
results are reminiscent of the mammalian NADases (ED
3\2\2\6), which, in their vast majority, also hydrolyse NADP
+
and pyridinium analogues of NAD
+
[17,34,35], as opposed to
most other NAD
+
-metabolizing enzymes, such as ADP-ribosyl-
transferases, which recognize only NAD
+
as substrate.
Alkaline inactivation
A striking feature that discriminates mammalian NADases of
different origins is the occurrence of paracatalytic inactivation
under alkaline conditions [36,37]. Thus, certain NADases, e.g.
from mouse, rat and rabbit, when incubated at pH 8 or higher,
undergo irreversible inactivation in the presence of NAD
+
,
whereas at more acidic pH values the glycohydrolase reaction
goes to completion [37]. As shown in Figure 1(A), human CD38
undergoes such self-inactivation at alkaline pH. Although the
hydrolysis of NAD
+
is practically linear with time at pH 6±5 and
7±5, it decreases progressively at pH 8±5. This was not due to
product inhibition of the reaction, because the addition of fresh
enzyme restored the reaction with a similar initial velocity (Figure
1A, arrow). Moreover, preincubation of the enzyme at pH 8±5
for 30 min in the absence of NAD
+
had no effect on its subsequent
activity (results not shown). Therefore the inactivation of CD38
is due to both NAD
+
turnover and alkaline pH and thus,
according to the Green and Dobrjansky classification [37], this
enzyme belongs to the category of the ‘self-inactivating’
Table 1 Kinetic characteristics of CD38 towards NAD
+
and analogues
The maximal rates are given as a percentage of the V
max
observed for NAD
+
(100% ¯ 0±3 nmol/min with an immunoprecipitate obtained from 10
6
Daudi cells). Abbreviations : ac
3
PyAD
+
,
3-acetylpyridine adenine dinucleotide; n.d., not detected.
Substrates … β-NAD
+
NADP
+
ε-NAD
+
ac
3
PyAD
+
α-NAD
+
cADPR NGD
+
K
m
(µM) 46³4 (8) 65³19 (7) 7³1 (6) 851³37 (4) – 224³9 (4) 1±6³0±4 (5)
Relative rate (%) 100 70 29 95 n.d. 16 39
Figure 1 Self-inactivation and INH sensitivity of CD38
(A) The NADase activity of immunopurified CD38 was assayed fluorometrically at 37 °C with
100 µM ε-NAD
+
in buffer A (see the text) adjusted to pH 6±5, 7±5or8±5, as indicated. In buffer
A, Hepes was replaced by 50 mM Pipes (pH 6±5) or 50 mM Tris (pH 7±5 and 8±5). At pH
8±5 an identical quantity of enzyme was added at 55 min (arrow). (B) Immunopurified CD38
was incubated at 37 °C in buffer A with 200 µM NAD
+
in the absence (trace a) or in the
presence (traces b and c) of 25 mM INH. HPLC analyses were performed on aliquots taken
at 0 h (trace b) or 3 h (traces a and c). The transglycosidation reaction product (elution time
5±2 min) was eluted with authentic hy
4
PyAD
+
[35]. Inset: progress curves of percentage NAD
+
disappearance (E) and hy
4
PyAD formation (D). For comparison the progress curve is given
for the disappearance of NAD
+
in the absence of INH (¬).
NADases. Recently, Han et al. [36] gave some convincing
evidence in an erythrocyte NADase, i.e. an NADase that operates
via a distinct mechanism [17], that this phenomenon might be
due to auto-ADP-ribosylation. CD38 was indeed found to be
ADP-ribosylated [38], but this modification could not be corre-

1386 V. Berthelier and others
Figure 2 Slow-binding inhibition of CD38 by araF-NAD
+
Assays were run fluorometrically at 37 °C in buffer A containing 20 µM ε-NAD
+
and 0–0±5
µM araF-NAD
+
. The representative progress curves show the increase in fluorescence at
410 nm (F®F
0
) observed after the addition of CD38 to an assay run in the absence (control) or
presence of 0±25 µM inhibitor. The analysis of such curves, by fitting them to eqn. (1) (see
the Experimental section), allowed the calculation of k. Inset, plot of k against inhibitor
concentration yielding k
off
(intercept) and k
on
, which can be calculated from the slope [see eqn.
(2)].
lated with its alkaline inactivation (V. Berthelier and P. Deterre,
unpublished work).
Slow-binding inhibition of CD38 by araF-NAD
+
One of the most potent inhibitors known for mammalian
NADases is araF-NAD
+
, an arabino analogue of NAD
+
, which
behaves as a reversible slow-binding inhibitor [32]. This molecule
showed similar inhibitory characteristics when tested on CD38.
As shown in Figure 2, when the assay was started with the
enzyme, the presence of araF-NAD
+
caused a progressive
increase in inhibition until a steady state was reached asympto-
tically. The transient pre-steady state was of the order of minutes.
Analysis of the kinetic data (Figure 2, inset), gave the following
results: k
on
¯ (10±20³0±01)¬10
%
M
−
"
[s
−
"
(association rate), k
off
¯ (1±72³0±05)¬10
−
%
s
−
"
(dissociation rate), t
"
#
¯ 67³2 min
(half-life of the E\I complex) and K
i
¯ 1±69³0±05 nM (n ¯ 3).
These results are in the range of those found for the inhibition of
bovine NADase by araF-NAD
+
[32]. A similar kinetic mechanism
can be proposed for CD38 that favours a slow interconversion
between E and E[I, the enzyme–inhibitor complex, as opposed
to a mechanism involving a slow transconformation of E[I into
a tighter final complex, E[I* [32].
The arabino analogues of NAD
+
are the only documented
inhibitors of NADases that give this unusual inhibition pattern;
the similarity of inhibition found for CD38 is therefore very
striking and highly significant in terms of likeness of reaction
mechanism and active-site topologies between the two classes of
enzyme. Moreover, araF-NAD
+
is, to our knowledge, the most
powerful inhibitor so far described for CD38 and should be an
interesting molecular tool to modulate its cellular activity.
Transglycosidation reactions catalysed by CD38 and kinetic
mechanism
One of the best characterized features of mammalian NADases
is the ability of most of these enzymes to catalyse a pyridine-base
exchange reaction [17]. Pyridinium NAD
+
analogues are pro-
Figure 3 Inhibition of CD38 NADase activity by reaction products
(A) Assays were run with CD38 from Daudi cells, at 37 °C, in buffer A containing various
concentrations of NAD
+
, in the presence of 0 (+), 1 (_)or3(E) mM nicotinamide. Double-
reciprocal plots of the NADase activity assayed by HPLC are shown. Similar results were
obtained with recombinant CD38 assayed fluorimetrically. Insert, cADPR hydrolase activity was
assayed under the same conditions. Symbols: +, control ; _, 1 mM nicotinamide ; E,2
mM nicotinamide. Ordinate unit, 10 nmol
−
1
[min; abscissa unit, 10 mM
−
1
.(B) Assays were run
with CD38 from Daudi cells, at 37 °C, in buffer A containing various concentrations of NAD
+
,
in the presence of 0 (+), 5 (_)or10(E) mM ADP-ribose. Double-reciprocal plots of the
NADase activity assayed by HPLC are shown. Similar results were obtained with recombinant
CD38 assayed fluorimetrically.
duced with retention of configuration and some can be used as
coenzymes for dehydrogenases [39]. We observed that CD38 was
capable of catalysing such transglycosidation reactions in the
presence of 3-acetylpyridine, 3-aminopyridine (results not shown)
and INH (Figure 1B). This last pyridine allows a classification of
NADases [40]. Some are inhibited by INH without leading to the
formation of the analogue and have been called ‘INH-sensitive’
[40]; in contrast, the ‘INH-insensitive’ NADases catalyse the
formation of 4-hydrazinocarbonylpyridine adenine dinucleotide
(hy
%
PdAD
+
). The fact that the rate of NAD
+
disappearance
remains unaffected in the presence of INH (Figure 1B, inset) and
that the analogue is formed, indicates that CD38 can be
categorized as an ‘ INH-insensitive’ NADase, as is the human
spleen NADase [40]. The ability of human CD38 to catalyse
transglycosidation with these different pyridines indicates that
this reaction might be relevant to the physiological role of CD38.
Our observation extends the work by Aarhus et al. [41] showing
that, under acidic pH, the transglycosidation of NADP
+
in the
presence of nicotinic acid yields nicotinic acid adenine dinucleo-
tide phosphate (NAADP
+
), a metabolite able to release intra-
cellular Ca
#
+
[42,43]. The occurrence of transglycosidation
reactions suggests a mechanism of action for CD38 that implies
the formation of an E[ADP-ribosyl reaction intermediary com-

1387Human CD38 is an authentic NAD(P)
+
glycohydrolase
Figure 4 NAD
+
methanolysis catalysed by CD38
Trace a, HPLC profile of the products of the non-enzymic solvolysis of NAD
+
[500 µM NAD
+
in buffer A, incubated at 90 °C and pH 5±0 for 90 min in the presence of 20 % (v/v) methanol].
The peaks were identified by co-elution with authentic α- and β-methyl ADP-ribose [28]. Trace
b, HPLC profile of the products of NAD
+
solvolysis catalysed by a CD38 immunoprecipitate in
the presence of 10% (v/v) methanol.
plex, resulting from the nicotinamide-ribose bond cleavage, that
can partition between different acceptors such as water and
pyridines. In agreement with this mechanism is the observation
that nicotinamide is a non-competitive inhibitor of CD38 with a
K
i
of 0±92³0±16 mM (n ¯ 4) and that ADP-ribose is a com-
petitive inhibitor with a K
i
of 4±2³0±4mM (n¯4) (Figure 3).
Together these results are similar to those found with classical
NADases, such as bovine spleen NADase [44]. As already
suggested [11], they are consistent with a minimal Ping Pong Bi
Bi kinetic mechanism for the reactions catalysed by CD38, in
which nicotinamide is the first reaction product released, and
which reduces to an ordered Uni Bi mechanism for hydrolysis
alone.
NAD
+
methanolysis reaction catalysed by CD38
An alternative acceptor for the ADP-ribosyl intermediate oc-
curring along the reaction pathway of NADases is methanol
[17,19,28,45], and it has been shown that this alcohol reacts in
competition with water leading to the formation of β-methyl
ADP-ribose [45]. As illustrated in Figure 4, CD38 was similarly
able to catalyse the methanolysis of NAD
+
and, by analogy with
the NADases studied so far, the reaction, which yielded the β-
form of methyl ADP-ribose, also occurred with an exclusive
retention of configuration.
We can estimate the relative efficacy of water and methanol in
the nucleophilic attack of the ADP-ribosyl intermediate by
calculating the partitioning ratio, K [28]:
K ¯ ([methyl ADP-ribose]}[ADP-ribose])¬([H
#
O]}[CH
$
OH])
In the presence of 2±5 M methanol, i.e. 10 % (v}v), this ratio was
12³1±7(n¯10). On a molar basis, methanolysis of NAD
+
catalysed by CD38 was therefore more than 10-fold faster than
hydrolysis: CD38 showed a clear selectivity for the solvent.
Importantly, this ratio was found to be constant throughout the
time course of the reaction and the extent of methanolysis was
strictly proportional to the concentration of methanol (results
not shown); this indicates that our experimental conditions were
not saturating and that the selectivity of methanol versus water
Table 2 Reaction products obtained by CD38-catalysed transformation of
NGD
+
in the presence of methanol
A is defined in eqn. (3). Results are means³S.D. for the number of replicates in parentheses.
Products (% of total)
[MeOH] (M) GDP-ribose Methyl GDP-ribose cGDPR A
013³1 (9) – 87³1 (9) 6±7³0±7 (9)
1±25 12³2 (9) 15³3 (9) 73³3 (9) 2±7³0±4 (9)
2±511³2 (9) 27³5 (9) 62³4 (9) 1±7³0±5 (9)
probably reflects more its intrinsic reactivity with the ADP-
ribosyl intermediate than a higher affinity for the active site of
the enzyme.
Mechanistic study of the ADP(GDP)-ribosyl cyclase reactions
Because of the small quantities of cADPR formed during NAD
+
hydrolysis catalysed by CD38 (see the Introduction section), it
was difficult to analyse the ADP-ribosyl cyclase activity of the
enzyme. In contrast, NGD
+
, which was shown to be a convenient
surrogate substrate to study the formation of cyclized compounds
[31,33], was converted in high yields (more than 80 % of reaction
products, Table 2) into cyclic GDP-ribose (cGDPR). The kinetic
parameters (K
m
and V
max
) measured (Table 1) were very similar
to those published previously [31,33].
To gain a better understanding of the molecular mechanism of
this cyclization process, we adopted a strategy that has proved
highly informative with bovine NADase: methanolysis [20]. By
analogy with NAD
+
(see above), CD38 was also able to catalyse
the methanolysis of NGD
+
; importantly, in the presence of
increasing concentrations of methanol, there was an increased
formation of methyl GDP-ribose, together with a net decrease in
cGDPR formation (Table 2). We excluded the possibility that
methyl GDP-ribose could arise from the CD38-catalysed sol-
volysis of cGDPR itself. As expected from its very low rate of
hydrolysis [33], we found that, under the same assay conditions,
the methanolysis of cGDPR remained negligible, forming less
than 3% of the whole NGD
+
-metabolizing activity (results not
shown). The partitioning ratio found for the methanolysis of
NGD
+
(48³7; n¯9) was higher than that found for NAD
+
(see
above) but was similar to that of bovine NADase [20]. The
competition between cyclase and hydrolysis}methanolysis ac-
tivities was quantified by the ratio, A of the different reaction
products:
A ¯ [cGDPR]}([GDP-ribose][methyl GDP-ribose]) (3)
As shown in Table 2, this ratio decreased with increasing
concentrations of methanol ; i.e. methyl GDP-ribose was pro-
duced by CD38 at the expense of cGDPR. This result indicates
clearly that the solvolysis and cyclization reactions are in
competition for a common intermediate and that cGDPR cannot
be a reaction intermediate in the transformation of NGD
+
into
GDP-ribose.
These results indicate that the molecular reaction mechanism
recently proposed for bovine spleen NADase [20] also applies to
CD38: i.e. a common ADP (GDP)-ribosyl intermediate formed
after the substrate’s nicotinamide-ribosyl bond cleavage gives
rise to the different reaction products, ADP(GDP)-ribose,
methyl-ADP(GDP)-ribose, cADPR (cGDPR) and pyridinium
analogues. The difference in yield in the formation of cADPR
and cGDPR can be attributed to the intrinsic reactivity (nucleo-

1388 V. Berthelier and others
philicity and positioning) of the purine N-positions (N-1 and N-
7 respectively) involved in the cyclization reactions within the
E[ADP(GDP)-ribosyl complexes [20].
Study of the cADPR hydrolase activity
Confirming earlier results [6,9,10], CD38 is also a cADPR
hydrolase. However, cADPR is a relatively poor substrate of
CD38 compared to NAD
+
(Table 1). This is also consistent with
the conclusion reached above. Indeed, as expected of a mech-
anism that assumes that cADPR is an obligatory reaction
intermediate in the conversion of NAD
+
into ADP-ribose, NAD
+
and cADPR should act as competing substrates for the enzyme.
The present finding, which indicates that the specificity ratio
V}K
m
is in favour of NAD
+
by a factor of 30 (estimated from the
results in Table 1), predicts that the cyclic metabolite should
accumulate during the major part of the reaction course. The fact
that cADPR remains a minor reaction product throughout the
conversion of NAD
+
into ADP-ribose [6,8–10] rules out the
possibility that the NADase activity of CD38 results from a
sequential ADP-ribosyl cyclase}cADPR hydrolase activity. A
similar conclusion was previously reached with bovine spleen
NADase [19,20].
As with NAD
+
, the incubation of cADPR with CD38 in the
presence of methanol also yielded β-methyl ADP-ribose (results
not shown). Moreover, a partitioning ratio of 14³1±4(n¯3)
was found, which is very close to that obtained with NAD
+
.
These results therefore suggest that the hydrolytic transformation
of NAD
+
and cADPR give rise to a common enzyme-stabilized
ADP-ribosyl reaction intermediate that can be trapped by
acceptors such as water, methanol or pyridines. This conclusion
is validated by the observation that nicotinamide is, as observed
with NAD
+
(see above), a non-competitive inhibitor of the
hydrolysis of cADPR (K
i
¯ 3³1mM;n¯4) (Figure 3A, inset)
and by the known ability of CD38 to catalyse the formation of
β-NAD
+
in the presence of cADPR and nicotinamide [11]. This
latter reaction is equivalent to the transglycosidation of NAD
+
,
where free nicotinamide can also react with the reaction in-
termediate to regenerate NAD
+
.
DISCUSSION
This work demonstrates that CD38 presents many of the catalytic
properties that characterize the well-known ‘classical ’ NADases:
transglycosidation, hydrolysis and methanolysis of NAD
+
with
retention of configuration, slow-binding inhibition by araF-
NAD
+
, ADP(GDP)-ribosyl cyclase and cADPR hydrolase ac-
tivities (summarized in Table 3). These features contrast strikingly
with the reactions catalysed by other NAD
+
-metabolizing en-
zymes such as mono-ADP-ribosyl transferases and poly(ADP-
ribose) polymerase, which are unable to catalyse the methanolysis
of NAD
+
and which transfer the ADP-ribosyl moiety to their
acceptors with inversion of configuration [46]. Although some
differences exist between the bovine spleen enzyme, which is the
best characterized mammalian NADase and the human CD38
molecule (for example the magnitude of the methanol-par-
titioning ratio and the kinetic parameters of ε-NAD
+
), they fall
within the normal range of differences found in this heterogeneous
family of enzymes. We therefore propose that CD38 belongs to
the EC 3\2\2\6 class of enzymes, the NAD(P)
+
nucleosidases,
which, in contrast with the EC 3\2\2\5 nucleosidases, hydro-
lytically cleave both NAD
+
and NADP
+
. To alleviate the
confusion that is pervasive in the nomenclature of the
NAD(P)
+
glycohydrolase, it should be emphasized that CD38
belongs to a family of enzymes very distinct from another class
Table 3 Selected CD38 enzymic properties compared with those of
classical mammalian NADases
Human CD38 NADases
Hydrolysis of NADP
+
? Yes Yes [17]
Hydrolysis of ε-NAD
+
? Yes Yes [17]
Transglycosidation with 3-aminopyridine
and 3-acetyl pyridine?
Yes Yes [17,39]
Sensitivity to INH ‘INH-insensitive ’ ‘INH-insensitive ’ (e.g. human)
and ‘INH-sensitive ’ [40]
Self-inactivation at alkaline pH (in the
presence of NAD
+
)?
Yes Yes/no [17]
Inactivation by thiol-reducing agents? Yes [48,49] Yes/no [17]
Inhibition by nicotinamide Non-competitive Non-competitive [17]
Inhibition by ADP-ribose Competitive Competitive [17,44]
Kinetic mechanism Ping Pong Bi Bi Ping Pong Bi Bi [44]
Inhibition by araF-NAD
+
Slow-binding Slow-binding [32]
Methanolysis of NAD
+
? Yes (retention of
configuration)
Yes (retention of configuration)
[17,45]
Formation of cADPR? Yes [6,8–10] Yes [18,20]
Hydrolysis of cADPR? Yes [6,8–10] Yes [18,19]
of NADases, mostly of microbial origin (e.g. Neurospora crassa),
which hydrolyse only NAD
+
and do not catalyse transglyco-
sidation reactions [17].
In the classification of mammalian NADases, human CD38}
NADase belongs to the ‘INH-insensitive ’ and ‘self-inactivating’
category. This probable identity between CD38 and classical
NADase is strengthened by the recent finding (D. Cockayne,
personal communication) that in CD38
−/−
knockout mice the
tissues that are traditionally considered to be the richest sources
of NADase, such as spleen, brain and liver [17], are totally
devoid of this enzyme activity.
The similarity between CD38 and bovine spleen NADase
allows one to draw some important conclusions about the
molecular mechanism of CD38 action. This enzyme, after the
cleavage of the nicotinamide-ribose bond of NAD
+
, generates an
E[ADP-ribosyl intermediate that can then be partitioned be-
tween different acceptors, i.e. water (hydrolysis), methanol
(methanolysis) and pyridines (transglycosidation). The chemical
nature of this intermediate, which in bovine spleen NADase
is most probably an oxocarbenium ion [28], remains to be
established for CD38. In this respect, the lower apparent select-
ivity for methanol versus water found in the solvolysis of this
intermediary complex could indicate, in the reactions catalysed
by CD38, a lesser oxocarbenium ion-like character [28]. However,
so far, bovine spleen NADase remains the only mammalian
NADase for which such an analysis is available ; further work is
also needed to assess this feature in CD38 of other species.
One of the most important issues that has been solved in this
study pertains to the multifunctionality of CD38. It has often
been indicated that the NADase activity of CD38 could in fact
result from a sequential combination of ADP-ribosyl cyclase
and cADPR hydrolase activities [6,8,10,16,33], and hence that
cADPR is the reaction intermediate in the hydrolytic conversion
of NAD
+
into ADP-ribose. We have clearly shown that this is
not true and that, in this respect, human CD38 functions very
similarly to bovine NADase: cADPR is a reaction product
whose poor yield is not due to its fast hydrolysis but to the low
reactivity, within the active site of CD38, of the N-1 position of
the adenine moiety ring of the ADP-ribosyl intermediate (see
also [20]). The final reaction scheme catalysed by CD38 is
summarized in Scheme 1.

1389Human CD38 is an authentic NAD(P)
+
glycohydrolase
Scheme 1 Reaction mechanism of CD38
Human CD38 catalyses the cleavage of the nicotinamide-ribose bond of NAD
+
(NGD
+
) leading to the formation of a E[ADP(GDP)-ribosyl intermediary complex. This intermediate can then partition
depending on competing reactions: (1) intermolecular reactions with acceptors such as water (hydrolysis), methanol (methanolysis) or pyridines (transglycosidation), and (2) an intramolecular reaction
between position N-1 of adenine (or N-7 of guanine when NGD
+
is used as substrate) and C-1« of the ribosyl moiety, yielding the cyclized products cADP-ribose (cGDP-ribose).
Finally, the finding that CD38 is most probably a classical
NADase has an important bearing on the investigation of its
physiological function, including recycling extracellular nucleo-
tides [47] or signalling functions [2–4]. It must take into account
the important and singular structural and catalytic heterogeneity
and the diversified tissue distribution that characterize this class
of enzymes.
We thank Dr. J. Banchereau and Dr. F. Brie
'
re (Schering Plough, Dardilly, France) for
kindly providing recombinant CD38 ; Dr. L. Boumsell for providing BB51 ascites
cells ; Professor N. J. Oppenheimer (University of California at San Francisco, San
Francisco, CA, U.S.A.) for his gift of araF-NAD
+
; and Dr. M. Muzard (University of
Reims, France) for her participation in some experiments. This work was supported
by grants from by the Centre National de la Recherche Scientifique, the Agence
Nationale de Recherche contre le SIDA, the Association pour la Recherche sur le
Cancer and the Ligue Nationale contre le Cancer. V. B. was supported by a fellowship
from the Direction de la Recherche et de la Technologie.
REFERENCES
1 Jackson, D. G. and Bell, J. I. (1990) J. Immunol. 144, 2811–2815
2 Malavasi, F., Funaro, A., Roggero, S., Horenstein, A., Calosso, L. and Mehta, K.
(1994) Immunol. Today 15, 95–97
3 Mehta, K., Shahid, U. and Malavasi, F. (1996) FASEB J. 10, 1408–1417
4 Lund, F., Solvason, N., Grimaldi, J. C., Parkhouse, R. M. E. and Howard, M. (1995)
Immunol. Today 16, 469–473
5 States, D. J., Walseth, T. F. and Lee, H. C. (1992) Trends Biochem. Sci. 17, 495
6 Howard, M., Grimaldi, J. C., Bazan, J. F., Lund, F. E., Santos-Argumedo, L.,
Parkhouse, R. M. E., Walseth, T. F. and Lee, H. C. (1993) Science 262, 1056–1059
7 Lee, H. C., Zoochi, E., Guida, L., Franco, L., Benatti, U. and De Flora, A. (1993)
Biochem. Biophys. Res. Commun. 191, 639–645
8 Summerhill, R. J., Jackson, D. G. and Galione, A. (1993) FEBS Lett. 335, 231–233
9 Takasawa, S., Tohgo, A., Noguchi, N., Koguma, T., Nata, K., Sugimoto, T., Yonekura,
H. and Okamoto, H. (1993) J. Biol. Chem. 268, 26052–26054
10 Zocchi, E., Franco, L., Guida, L., Benatti, U., Bargellesi, A., Malavasi, F., Lee, H. C.
and De Flora, A. (1993) Biochem. Biophys. Res. Commun. 196, 1459–1465
11 Inageda, K., Takahashi, K., Tokita, K., Nishina, H., Kanaho, Y., Kukimoto, I., Kontani,
K., Hoshino, S. and Katada, T. (1995) J. Biochem. (Tokyo) 117, 125–131
12 Galione, A. (1993) Science 259, 325–326
13 Lee, H. C. (1994) Mol. Cell. Biochem. 138, 229–235
14 Sitsapesan, R., McGarry, S. J. and Williams, A. J. (1995) Trends Pharmacol. Sci. 16,
386–391
15 Kato, I., Takasawa, S., Akabane, A., Tanaka, O., Abe, H., Takamura, T., Suzuki, Y.,
Nata, K., Yonekura, H., Yoshimoto, T. and Okamoto, H. (1995) J. Biol. Chem. 270,
30045–30050
16 Zocchi, E., Franco, L., Guida, L., Piccini, D., Tacchetti, C. and De Flora, A. (1996)
FEBS Lett. 396, 327–332
17 Price, S. R. and Pekala, P. H. (1987) in Pyridine Nucleotide Coenzymes : Chemical,
Biochemical and Medical Aspects, (Dolphin, D., Poulson, R. and Avramovic, O., eds.),
pp. 513–548, John Wiley & Sons, New York
18 Kim, H., Jacobson, E. L. and Jacobson, M. K. (1993) Science 261, 1330–1333
19 Muller-Steffner, H., Muzard, M., Oppenheimer, N. and Schuber, F. (1994) Biochem.
Biophys. Res. Commun. 204, 1279–1285
20 Muller-Steffner, H. M., Augustin, A. and Schuber, F. (1996) J. Biol. Chem. 271,
23967–23972
21 Koguma, T., Takasawa, S., Tohgo, A., Karasawa, T., Furuya, Y., Yonekura, H. and
Okamoto, H. (1994) Biochim. Biophys. Acta 1223, 160–162
22 Mizuguchi, M., Otsuka, N., Sato, M., Ishii, Y., Kon, S., Yamada, M., Nishina, H.,
Katada, T. and Ikeda, K. (1995) Brain Res. 697, 235–240
23 Randall, T. D., Lund, F. E., Howard, M. C. and Weissman, I. L. (1996) Blood 87,
4057–4067
24 Kramer, G., Steiner, G., Fo
$
dinger, D., Fiebiger, E., Rappersberger, C., Binder, S.,
Hofbauer, J. and Marberger, M. (1995) J. Urol. 154, 1636–1641
25 Alessio, M., Roggero, S., Funaro, A., De Monte, L. B., Peruzzi, L., Geuna, M. and
Malavasi, F. (1990) J. Immunol. 145, 878–884
26 Deaglio, S., Dianzani, U., Horenstein, A. L., Ferna
'
ndez, J. E., Van Kooten, C.,
Bragardo, M., Funaro, A., Garbarino, G., Di Virgilio, F., Banchereau, J. and Malavasi,
F. (1996) J. Immunol. 156, 727–734
27 Boumsell, L., Schmid, M., Dastot, H., Gouttefangeas, C., Mathieu-Mahul, D. and
Bensussan, A. (1990) J. Immunol. 145, 2797–2802
28 Tarnus, C., Muller, H. M. and Schuber, F. (1988) Bioorg. Chem. 16, 38–51
29 Muller, H. M., Muller, C. D. and Schuber, F. (1983) Biochem. J. 212, 459–464
30 Graeff, R. M., Mehta, K. and Lee, H. C. (1994) Biochem. Biophys. Res. Commun.
205, 722–727
31 Graeff, R. M., Walseth, T. F., Hill, H. K. and Lee, H. C. (1996) Biochemistry 35,
379–386
32 Muller-Steffner, H. M., Malver, O., Hosie, L., Oppenheimer, N. J. and Schuber, F.
(1992) J. Biol. Chem. 267, 9606–9611
33 Graeff, R. M., Walseth, T. F., Fryxell, K., Branton, W. D. and Lee, H. C. (1994) J. Biol.
Chem. 269, 30260–30267
34 Schuber, F., Travo, P. and Pascal, M. (1979) Bioorg. Chem. 8, 83–90
35 Tarnus, C. and Schuber, F. (1987) Bioorg. Chem. 15, 31–42
36 Han, M. K., Lee, J. Y., Cho, Y. S., Song, Y. M., An, N. H., Kim, H. R. and Kim, U. H.
(1996) Biochem. J. 318, 903–908
37 Green, S. and Dobrjansky, A. (1971) Biochemistry 10, 4533–4538
38 Grimaldi, J. C., Balasubramanian, S., Kabra, N. H., Shanafelt, A., Bazan, J. F.,
Zurawski, G. and Howard, M. C. (1995) J. Immunol. 155, 811–817
39 Anderson, B. M. (1982) in The Pyridine Nucleotide Coenzymes, (Everse, J., Anderson,
B. & You, K. S., eds.), pp. 91–133, Academic Press, New York
40 Zatman, L. J., Kaplan, N. O., Colowick, S. P. and Ciotti, M. M. (1954) J. Biol. Chem.
209, 453–466
41 Aarhus, R., Graeff, R. M., Dickey, D. M., Walseth, T. F. and Lee, H. C. (1995) J. Biol.
Chem. 270, 30327–30333
42 Chini, E. N. and Dousa, T. P. (1995) Biochem. Biophys. Res. Commun. 209,
167–174

1390 V. Berthelier and others
43 Lee, H. C. and Aarhus, R. (1995) J. Biol. Chem. 270, 2152–2157
44 Schuber, F., Travo, P. and Pascal, M. (1976) Eur. J. Biochem. 69, 593–602
45 Pascal, M. and Schuber, F. (1976) FEBS Lett. 66, 107–109
46 Oppenheimer, N. J. and Handlon, A. L. (1992) in The Enzymes, (Sigman, D. S., ed.),
pp. 453–505, Academic Press, San Diego
47 Deterre, P., Gelman, L., Gary-Gouy, H., Arrieumerlou, C., Berthelier, V., Tixier, J. M.,
Received 28 August 1997/17 November 1997; accepted 12 December 1997
Ktorza, S., Goding, K., Schmitt, C. and Bismuth, G. (1996) J. Immunol. 157,
1381–1388
48 Kontani, K., Nishina, H., Ohoka, Y., Takahashi, K. and Katada, T. (1993) J. Biol.
Chem. 268, 16895–16898
49 Franco, L., Zocchi, E., Calder, L., Guida, L., Benatti, U. and De Flora, A. (1994)
Biochem. Biophys. Res. Commun. 202, 1710–1715