Research Article

Changes in red cell ion transport, reduced intratumoral

neovascularization, and some mild motor function abnormalities
accompany targeted disruption of the Mouse Kell gene (Kel)

Xiang Zhu,1 Alicia Rivera,2 Mari S. Golub,3 Jianbin Peng,1 Quan Sha,1 Xu Wu,1 Xiaoling Song,1
Prem Kumarathasan,4 Mac Ho,5 Colvin M. Redman,1 and Soohee Lee1

Kell (ECE-3), a highly polymorphic blood group glycoprotein, displays more than 30 antigens that produce
allo-antibodies and, on red blood cells (RBCs), is complexed through a single disulfide bond with the inte-
gral membrane protein, XK. XK is a putative membrane transporter whose absence results in a late onset
form of neuromuscular abnormalities known as the McLeod syndrome. Although Kell glycoprotein is known
to be an endothelin-3-converting enzyme, the full extent of its physiological function is unknown. To study
the functions of Kell glycoprotein, we undertook targeted disruption of the murine Kel gene by homologous
recombination. RBCs from Kel(–/–) mice lacked Kell glycoprotein, Kell/XK complex, and endothelin-3-con-
verting enzyme activity and had reduced levels of XK. XK mRNA levels in spleen, brain, and testis were
unchanged. In Kel(–/–) mice RBC Gardos channel activity was increased and the normal enhancement by
endothelin-3 was blunted. Analysis of the microvessels of tumors produced from LL2 cells indicated that the
central portion of tumors from wild-type mice were populated with many mature blood vessels, but that ves-
sels in tumors from Kel(–/–) mice were fewer and smaller. The absence of Kell glycoprotein mildly affected
some motor activities identified by foot splay on the drop tests. The targeted disruption of Kel in mouse
enabled us to identify phenotypes that would not be easily detected in humans lacking Kell glycoprotein. In
this regard, the Kell knockout mouse provides a good animal model for the study of normal and/or patho-
physiological functions of Kell glycoprotein. Am. J. Hematol. 84:492–498, 2009. V C 2009 Wiley-Liss, Inc.

Introduction bioactive endothelin-3 [8,9]. Substance P and neurokinin A

In humans, Kell glycoprotein is a highly polymorphic 93 kDa,
type II membrane glycoprotein [1]. More than 30 different anti-
gens that have been shown to produce allo-antibodies are
associated with Kell glycoprotein [2]. Mutations of the Kell
gene resulting in a Kell null phenotype are known, and about
15 null alleles have been reported [3]. Kell null phenotypes are
considered normal and no clinical abnormalities associated
with null alleles have been documented. On red cells, Kell gly-
coprotein is linked by a single disulfide bond to its partner pro-
tein, XK, a putative membrane transporter that is predicted to
traverse the membrane 10 times and expresses the single
antigen, Kx, that produces the alloantibody, anti-Kx [4–6]. The
expression of Kell glycoprotein and XK are each affected by
the absence of the other partner. McLeod red cells, which lack
XK, are characteristically acanthocytic, but are also associated
with severe depression of all Kell antigens. Red blood cells
(RBCs) from Kell null phenotypes also have less XK. A notice-
able characteristic of Kell null red cells is that, although these
cells have less XK protein, the Kx antigen, as detected sero-
logically, increases. It has been speculated that normally the
Kell glycoprotein partially covers the Kx epitope, but that when
Kell glycoprotein is absent, Kx is fully exposed [2,7].
Kell glycoprotein is a member of the M13 family of zinc
endopeptidases which includes Kell, neutral endopeptidase
24.11 (NEP), two endothelin-converting enzymes (ECE-1
and ECE-2), the product of the PEX gene, and XCE, a pro-
tein preferentially expressed in the central nervous system.
A common functional feature of the M13 family is that all of
the members are involved in the activation, or inactivation,
of bioactive peptides. Unlike ECE-1 and ECE-2 which have
a preference for big endothelin-1, Kell glycoprotein has a
preference for big endothelin-3, which it cleaves to produce
have also recently been shown to be cleaved by Kell, but
its affinities for these substrates are relatively low [10].
Endothelins have multiple biological functions, and oper-
ate through the activation of two G-protein-coupled recep-
tors, endothelin receptor A (ETA) and endothelin receptor
B (ETB), to regulate vasoconstriction, vasodilation, and cell
proliferation [11–14]. Endothelin-1 and endothelin-2 binds
both ETA and ETB while endothelin-3 preferentially binds to
ETB mediating the release of nitric oxide (NO) and prosta-
Additional Supporting Information may be found in the online version of this
article.
1
Department of Pathology, New York Blood Center, New York, New York;
2
Harvard Medical School, Children’s Hospital Boston, Boston, Massachu-
setts; 3California National Primate Research Center, University of California–
Davis, Davis, California; 4Environmental Health Centre, Ottawa, Ontario,
Canada; 5Harvard Medical School, Massachusetts General Hospital, Char-
lestown, Massachusetts
Contract grant sponsor: National Institute of Health, a Specialized Center of
Research (SCOR) grant in Transfusion Biology and Medicine, HL54459 and
an NIH grant, 5R01 HL075716.
Conflict of interest: Nothing to report.
Xiang Zhu is currently at Peking University Health Science Center, Beijing,
People’s Republic of China.
Quan Sha is currently at Immunomedics, Inc., Morris Plains,New Jersey.
Mac Ho is currently at National Cancer Center, Singapore.
*Correspondence to: Soohee Lee, The New York Blood Center, 310 East
67th Street, New York, NY 10065. E-mail: solee@nybloodcenter.org
Received for publication 13 May 2009; Accepted 13 May 2009
Am. J. Hematol. 84:492–498, 2009.
Published online 18 May 2009 in Wiley InterScience (www.interscience.wiley.
com).
DOI: 10.1002/ajh.21453

V
C 2009 Wiley-Liss, Inc.

American Journal of Hematology 492 http://www3.interscience.wiley.com/cgi-bin/jhome/35105


Figure 1. Absence of Kell transcripts in spleen and testis, and absence of Kell
glycoprotein and endothelin-3-converting activity in RBCs of Kel(–/–) mice com-
pared with WT mice. A: Northern blot of spleen and testis. B: Western blot of
RBCs. C: Endothelin-3-converting activity of RBCs. bET-1, bET-2, and bET-3
denote big ET-1, big ET-2, and big ET-3, respectively. Data are expressed as
means ± SD of three sets of separate experiments.
cyclin involving cGMP pathway [15–17]. Endothelin-3 also
functions in neovascularization [18], migration of endothelial
cells [19,20], guidance of axonal growth [21], promotion of
cancers [22–24], and developmental processes of enteric
neurons and epidermal melanocytes, affecting migration
and differentiation of neural crest-derived cells [25].
Targeted disruption of mouse ETB revealed that homozy-
gous mutants, deficient in ETB, had a phenotype identical
to human Hirschsprung’s disease which is associated with
defects in the endothelin-3 gene or in ETB. These mice dis-
played aganglionic megacolon and changes in coat color
with abnormal epidermal melanocytes. Mutant mice defi-
cient in endothelin-3 had a similar phenotype, suggesting
that endothelin-3 is the natural ligand for the ETB receptor
during development [26,27]. Two naturally occurring, reces-
sive mutant mice, piebald-lethal and lethal spotting which
have gene defects in ETB and endothelin-3 respectively,
were found to display a phenotype identical to that of ETB
and ET-3 knockout mice [28,29]. ECE-1 deficient mice also
possessed a phenotype similar to ETB and ET-3 deficient
mice, indicating that big ET-3 is an important substrate for
ECE-1 even though ECE-1 has a preference for big ET-1
and big ET-2. Thus, ECE-1 has overlapping function with
Kell glycoprotein [26]. Kell null individuals are not known to
exhibit disease symptoms associated with defects in endo-
thelin-3 or in ETB, although Kell glycoprotein has a prefer-
ence for big endothelin-3. Thus, Kell glycoprotein may not
replace the function of ECE-1 in generating Hirschsprung’s
disease or other related phenotypes. The expression of Kell
glycoprotein is largely limited to erythroid tissues where it is
American Journal of Hematology
research article
Figure 2. Reduction of XK protein in red cells and unchanged XK mRNA level in
testis, brain, and spleen of Kel(–/–) mice. A: Reduced level of XK protein, Kel(–/–)
lanes of right panel (first 4 lanes, reduced Western blot) and absence of Kell/XK
complex in the Kel(–/–) lanes (last 4 lanes, nonreduced Western blot). Left and mid-
dle panels are blots stained by Ponceau S to show even loading of the sample pro-
teins before probing the blot by anti-XK (Left 4-lane-panel is reduced and the middle
4-lane-panel is nonreduced samples). The samples are red blood cell ghosts from
two WT and two Kel(–/–) mice. B: XK mRNA expression in testis, brain, and spleen
of Kel(–/–) and WT mice by Northern blot analysis. Upper panel is probed with 32P-
labeled XK cDNA and the lower panel is probed with 32P-labeled b-actin cDNA.
linked to XK protein and where ECE-1, a homodimer, is not
present [8]. Although no clinical abnormalities associated
with Kell null alleles have been documented, the full extent
of its physiological function is unknown [3]. The expression
of Kell glycoprotein in nonerythroid tissues in human and in
mouse might be different and only minor amounts in speci-
alized cell types may express Kell glycoprotein. [1,30–32].
Mouse Kell (mKell) [33] is a 110 kDa type II membrane gly-
coprotein having 80% nucleotide and 74% amino acid
sequence identity with human Kell. The mouse Kell gene
(Kel) is located on chromosome 6, 20.5 cM and, like its
human counterpart, is a single copy gene organized into 19
exons. As in human red cells [6], mKell is disulfide-linked to
XK and has endothelin-3-converting enzyme activity. Com-
pared to human Kell glycoprotein, mKell has 8 N-linked
sugars rather than 5, an additional extracellular cysteine near
the transmembrane region, and a shorter (by 19 amino acids)
intracellular N-terminal domain. To study the physiological
function of Kell glycoprotein, we deleted, and substituted with
a neomycin cassette, a segment of Kel that encodes the zinc
binding enzyme active motif, HELLH. Here, we report the
generation of the targeted disruption of the mouse Kell gene
(Kel) and the biochemical and functional phenotype evalua-
tions of Kell glycoprotein deficient mice.
Results
Characterization of the Kel(–/–) gene
Targeted disruption of Kel was performed by standard
homologous recombination procedures as described in
493
research article
TABLE I. Blood Pressure and Pulse of Kel(–/–) and WT Mice at 12 Months of
age

Genotype Systolic ± SE Pulse ± SE Systolic success ± SE

Kel(–/–), N 5 6 100.3 ± 5.84 658.7 ± 61.0 15.33 ± 1.28

WT, N 5 6 109.1 ± 6.11 751.3 ± 16.8 18.17 ± 0.48
P 5 0.320a P 5 0.005a P 5 0.065a

a
t-test P-value at 0.05.

TABLE II. Red Cell Indices of Kel(–/–) and WT mice

Attribute WT ± SD N 5 25a Kel(–/–) ± SD N 5 17a P-valuea at 0.05

HGB (g/dL) 10.37 ± 0.83 10.85 ± 0.50 0.11

MCV (fL) 46.98 ± 0.69 48.43 ± 0.88 <0.0001
MCH (qg) 13.51 ± 0.39 14.23 ± 0.50 <0.0001
MCHC (g/dL) 28.74 ± 0.66 29.33 ± 0.81 0.024
Figure 3. RT-PCR of spinal cord of WT and Kel(–/–) mice.Lane 1, 1 kb DNA lad- CH (qg) 13.40 ± 0.20 13.81 ± 0.28 <0.0001
der; lanes 2–5, Kel(–/–); lanes 6–9, WT. Lanes 2 and 6 are for Kel RT-PCR HDW (g/dL) 2.05 ± 0.11 2.16 ± 0.15 0.013
(351bp) showing the presence of 351-bp band only in WT (arrow). Lanes 3 and 7 ReticP 2.93 ± 0.56 3.35 ± 0.60 0.064
are XK (373 bp). Lanes 4 and 8 are the positive control, G3PDH (456 bp). RBC (x103 cells/lL) 7.64 ± 0.57 7.65 ± 0.43 0.96
HCT (%) 35.78 ± 2.70 36.95 ± 1.90 0.23

Materials and Methods and summarized in Supporting Infor- a

mation Fig. 1. Examples of results of the screening are
shown in Supporting Information Fig. 2A. Heterozygous DNA
showed the presence of both bands for WT and Kel knockout
alleles when cut with BamHI or EcoRV (last lanes of A, upper
and lower panel). Of 5 chimeric mice, 2 exhibited germ line
transmission of the Kel targeted allele. Results of PCR
screening of genomic tail DNAs are illustrated in Supporting
Information Fig. 2B, which shows 349 bp for wild type (lane
2), 941 bp for Kel(–/–) (lane 5), and both bands for Kel(1/–)
(lanes 6 and 7).
Lack of Kell mRNA and Kell glycoprotein
in Kel(–/–) mice
Northern blot analysis of Kell mRNA in spleen and testis
demonstrated the absence of normal Kell mRNA (2.5 kb) in
Kel(–/–) mice (Fig. 1A). The low molecular weight bands
seen in Kel(–/–) mice indicate the presence of truncated
Kell mRNA (which would be detected by the probe used).
This was confirmed by the presence of the predicted RT-
PCR product derived from the nontargeted area of the Kel
gene (data not shown). Western blot analysis of RBC mem-
brane proteins from Kel(–/–) mice showed the absence of
Kell glycoprotein (Fig. 1B, right panel, second lane). As
shown in Fig. 1C, there was also a lack of endothelin-3-
converting enzyme activity by Kel(–/–) red cells. Heterozy-
gous Kel(1/–) red cells showed decreased activity
compared to wild-type red cells (Fig. 1C). Together, these
results demonstrated that the Kel knockout successfully
disrupted Kell glycoprotein expression and removed its
enzymatic activity.
Red cell XK protein is reduced but XK mRNA is
unchanged in Kel knockout mice
XK protein was decreased in Kel(–/–) red cells compared
to wild-type red cells (Fig. 2A, right panel, reduced western
blot, Kel(–/–) lane). Kell/XK complex is also absent from
Kel(–/–) RBCs (Fig. 2A, right panel, nonreduced, Kel(–/–)
lane). Panels stained with Ponceau S show even loading of
samples. XK protein is reduced in Kel knockout red cells as
shown in 2A, but mRNA levels are unchanged in testis and
spleen where Kell glycoprotein and XK are coexpressed
(Fig. 2B). The XK mRNA level in brain, where little or no
Kell glycoprotein is expressed, is also unchanged.
Expression of Kell in mouse spinal cord
Detection of expression of Kell in mouse brain has been
inconclusive and it has been suggested that there may be
Statistical significance of genotype effect was analyzed by two-way ANOVA
on the two sets of data: one set contained 18 WT and 9 Kel(–/–) mice, and the
second set contained 7 WT and 8 Kel(–/–) mice. HGB, hemoglobin concentration;
MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; MCHC,
mean corpuscular hemoglobin concentration; CH, red cell hemoglobin content;
HDW, hemoglobin distribution width; ReticP, reticulocyte percentage; RBC, red
blood cell count; HCT (%), hematocrit percentage.
species differences between mouse and human [31]. As
determined by in situ hybridization, the expression of
mouse Kell was high in erythroid, but not in nonerythroid
tissues. By RT-PCR, mouse Kell mRNA was detected in
testis but not in muscle or brain. However, using Kel(–/–)
tissue as a negative control, RT-PCR of spinal cord from
wild-type mouse indicated weak expression of Kell tran-
scripts in this tissue (Fig. 3). Lanes 5 and 9 (Fig. 3) show
that a weak RT-PCR band (235 BP) of an erythroid specific
gene product, glycophorin A (GPA), was present as well,
which may indicate contamination by erythroid cells, but
expression of GPA in nonerythroid cells is also possible as
shown in the GPA gene (GYPA) expression profiles avail-
able at GeneCards site [34–36].
Blood pressure, pulse, and ECG analysis
The systolic blood pressure and pulse of 6 WT and 6
Kel(–/–) mice were measured. As shown in Table I, systolic
blood pressure was not significantly different between WT
and Kel(–/–) mice, but pulse of Kel(–/–) mice was signifi-
cantly slower than among WT mice (P < 0.005). ECG anal-
ysis showed that the P-Q and P-R intervals were signifi-
cantly decreased in Kel(–/–) mice than in WT mice (P <
0.05) suggesting that there may be some effect coming
from altered sympathovagal balance probably involving
changes in NO signaling [37,38].
Hematological analysis of Kel(–/–) mouse blood
Two sets of analyses done on two different dates were
combined and analyzed by two-way ANOVA. There were
small but significant increases in MCV, MCH, CH, MCHC,
and HDW in Kel(–/–) RBCs compared to WT RBCs (Table
II). However, the biological significance of these small
increases is not known.
Measurement of channel activity of RBCs
and the effect of ET-3
Because Kell glycoprotein activates ET-3 and ET-3 is
known to modulate Ca11activated K1 channels (Gardos
channel) [39] it was important to investigate Gardos
channel activity in Kel(–/–)RBCs. Therefore, we measured

494 American Journal of Hematology

 research article
TABLE III. Counting of CD31 Positive Microvessels in the Central Part of
Tumors

Age MCV counts/ tumor WT ± SE (Na) Kel (–/–) ± SE (Na) t-test P-value

5 months 2-week tumor 31.42 ± 4.68 (12) 10.00 ± 1.60 (9) 0.0012b
3-week tumor 30.20 ± 6.52 (5) 17.88 ± 5.51 (8) 0.18
12 months 2-week tumor 31.33 ± 2.73 (3) 17.80 ± 3.40 (5) 0.034b
3-week tumor 23.75 ± 4.23 (4) 5.333 ± 4.33 (3) 0.031b

a
N 5 number of tumors counted.
b
P value  0.05.

Figure 4. Comparison of Gardos channel activity of RBCs of Kel(–/–) and WT

mice and the effect of ET-3. Gardos channel activity is defined by ChTx-sensitive
K1 influx measured by 86Rb uptake. Data are expressed as means ± SD of 5 to 7
separate experiments in duplicate determinations. Statistical analysis was eval-
uated by student’s t-test: *P < 0.05, n 5 5 to 7.

the maximal activity of Gardos Channel in freshly isolated

erythrocytes from Kel(–/–) mice compared to WT. We found
that the deletion of Kell glycoprotein induced an increase in
Gardos channel activity under maximal conditions (Fig. 4).
These results suggested that Kell glycoprotein is involved
in the regulation of the Gardos channel. To determine if the
regulatory effect of ET-3 on the Gardos channel, as
observed previously [39], was disrupted by Kell deletion,
we measured Gardos channel activity in the presence of
ET-3. As shown in Figure 4, the activation effect of ET-3 on
the Gardos channel was blunted by the deletion of Kell,
suggesting that Kell glycoprotein plays an important role in
the interaction of ET-3 with the Gardos channel, possibly by
facilitating receptor-ligand interactions or activation of the
signal transduction pathway involving ET-3 and ETB. It is
known that the functional ETB is present on RBCs [39,40].

Analysis of microvessel formation in Lewis lung

carcinoma tumors
In two sets of experiments, tumors grown in 5- and
12-month-old mice for 2 weeks and 3 weeks were col-
lected, weighed, and measured. There was no difference in
size and weight of tumors from WT and Kel(–/–) mice. The
number of microvessels at the boundary (periphery) of
tumors in WT and Kel(–/–) mice did not differ. However, the
number of microvessels, identified by CD31 immunostain-
ing, in the central part of 2-week tumors was significantly
lower in Kel(–/–) mice than in WT in both age groups (P 5
0.0012 for 5-month-old mice and P 5 0.0340 for 12-month-
old mice). In Kel(–/–) mice 3-week tumors from 12-month-
old mice exhibited significantly reduced numbers of Figure 5. Microvessels labeled by CD31 in the central part of 2-week tumors in
microvessels (P 5 0.031). Similar tumors from 5-month-old WT (A, C, E, and G ) and Kel(–/–) mice (B, D, F, and H ). Magnification: A, B x40;
mice exhibited reduced numbers of microvessels, but the C, D x100; E, F x200; G, H x400.
difference was not significant (P 5 0.1835) (Table III).
Examples of immunohistologic examination, shown in
Fig. 5, illustrate that in WT mice the central part of the Similarly, grip strength, stride length, and wire hang tests
tumors are rich in microvessels and the vessels appear to were not consistently affected by lack of Kell. However, a
be mature (Fig. 5A,C,E,G). In contrast, in Kel(–/–) mice, decline in forelimb strength was observed in Kel(–/–) mice
the microvessels in the central part of the tumors are fewer from 12 to 16 months of age P 5 0.014), but not in WT mice
and smaller (Fig. 5B,D,F,H). (P 5 0.55). Further, mice lacking Kell had greater hindlimb
foot splay in the drop test over all three time points (6, 12,
Mild motor dysfunction phenotype of Kel(–/–) mice and 16 months of age) by RMANOVA (P 5 0.012) (Table IV).
No severe neurological symptoms (stereotypy, seizures, or On the rotarod test, mice lacking Kell generally had more
autonomic signs) were recorded in the Functional Observa- falls during training and a shorter time to fall during testing at
tional Battery. Normal responses to handling, visual, tactile, increasing speeds, but the group effect approached signifi-
and auditory stimulation were seen. Scores for posture and cance only for the second trial at the highest speed (44 rpm)
the simple reflexes were in the normal range. However, in when the mean time to fall was 13 ± 3 sec in the WT mice
observational scoring of gait abnormality, 1 of 6 WT mice and 6 ± 3 in the Kel(–/–) mice ( P 5 0.054). There were no
received an abnormal score as compared to 3 of 6 Kel (–/–). effects of genotype on measures of spontaneous activity
This difference was not statistically significant. (horizontal and vertical movements or distance traveled).

American Journal of Hematology 495

research article
TABLE IV. Grip Strength, Wire Hang, Stride Length, and Foot Splay in the spinal cord, and may be expressed only in some special-
Drop Test of Kel(–/–) Mice Compared With Wild- Type Mice ized cells in nervous tissues.
Age Age
There was no significant difference in plasma endothelin-
3 levels between Kel(–/–) and WT mice (data not shown).
6 months 12 months 16 months 6 months 12 months 16 months On red cells, Kell glycoprotein may function as an endothe-
Grip Forelimb grip strength ± SD (g) Hindlimb grip strength ± SD (g)
lin-3-converting enzyme only when the substrate, big
endothelin-3, is available in sufficient concentration and the
WT 0.85 ± 0.80 ± 0.76 ± 0.59 ± 0.49 ± 0.49 ±
0.13 0.11 0.13 0.10 0.17 0.13
local pH is optimal for Kell enzymatic activity. The substrate
Kel(–/–) 0.93 ± 0.78 ± 0.72 ± 0.50 ± 0.51 ± 0.46 ± concentration (big endothelin-3) in plasma is in the picomo-
0.14 0.16 0.11 0.11 0.19 0.19 lar range, which is much less than the Km value, which is
Wire hang Time to fall ± SD (sec) Number of falls ± SD
in the micromolar range [8]. We hypothesize that Kell glyco-
protein, as an endothelin-3-converting enzyme, may have a
WT 27 ± 15 29 ± 24 21 ± 13 0.67 ± 1.33 ± 2.00 ± limited role in normal physiology, but may function when
0.82 0.51 0.89 red cells are carried to pathological sites, such as tumors,
Kel(–/–) 18 ± 6 30 ± 11 24 ± 5 1.33 ± 1.67 ± 1.67 ±
0.81 0.82 1.03 where there is a suitable microenvironment for its activa-
tion. Kell null individuals, like Kel(–/–) mice, have no easily
Stride Forelimb stride length ± SD (mm) Hindlimb stride length ± SD (mm) detected abnormalities and parameters such as lower heart
WT 64 ± 8 69 ± 7 66 ± 7 73 ± 6 68 ± 10 66 ± 7 rates and decreased motor activities. If present, such
Kel(–/–) 68 ± 5 71 ± 6 69 ± 7 72 ± 5 71 ± 4 71 ± 13 abnormalities and parameters may go undetected in a nor-
Drop Forelimb drop foot splay ± SD Hindlimb drop foot splay ± SD
mal clinical examination. When Kell glycoprotein is absent
(mm) (mm)a the amount of XK is reduced and the Kell/XK complex no
longer exists [4,6]. XK may play a role in regulating K1 flux
WT 16 ± 3 20 ± 3 18 ± 2 27 ± 4 32 ± 3 31 ± 2
Kel(–/–) 19 ± 3 19 ± 3 17 ± 2 32 ± 4b 39 ± 3c 41 ± 2d
and the reduction in XK in Kel(–/–) mice may account for
the difference in Gardos channel activity seen in Kel(–/–)
a
Significance of the effect of Kel genotype on hindlimb foot splay in the drop mice. It is also possible that XK function differs depending
test over all three time points (6, 12, 16 months) by RMANOVA, P 5 0.012; 6 on whether it exists in a complex with Kell glycoprotein or
month time point. as a monomer. In this respect, Kell glycoprotein may be a
b
P 5 0.40; 6 month time point. regulator of XK ion transport function.
c
P 5 0.036; 12 month time point.
d
The increased number of microvessels in the central part
P 5 0.021; 16 month time point.
of tumors in WT mice compared to Kel(–/–) mice indicates
that Kell glycoprotein may be involved in the formation of
Discussion blood vessels or in the interaction of vascular endothelial
Here, we report analysis of mice carrying targeted dis- cells with tumor cells. The observation that the center and
ruption of the Kel gene. We found that RBCs from Kel(–/–) not the periphery of the tumor raised in Kel(–/–) mice had
mice lacked Kell glycoprotein, Kell/XK complex, and endo- less microvessel formation suggests that in that microenvir-
thelin-3-converting enzyme activity, and had reduced levels onment, Kell, by it’s endothelin-3-converting enzyme activity,
of XK as is seen in Kell null humans [3]. XK mRNA levels is involved in neovascularization in a manner that is nonover-
in spleen, brain, and testis, however, were unchanged. In lapping with the function of ECE-1 [26]. We expect that the
Kel(–/–) RBCs, Gardos channel activity was increased but center of the tumor may be hypoxic and acidic due to anaero-
the modulation of the channel by endothelin-3 was blunted. bic glycolysis. The optimum pH for Kell enzyme activity is
The central portions of tumors from WT mice were popu- acidic, pH 6 [8], and in addition the Kell promoter has HIF1
lated with many mature blood vessels in comparison with (Hypoxia-Inducible Factor 1) binding sites (nt -829 and -
those of Kel(–/–) mice which were fewer and smaller. In 1655) although we do not know if they are functionally active.
behavioral tests, the absence of Kell glycoprotein mildly If they are, the expression of the Kell gene is hypoxia induci-
affected some motor activities identified by foot splay on ble. These conditions are beneficial for Kell enzymatic func-
the drop tests. tion and for activating endothelin-3 locally. As noted earlier,
Reduction of XK levels in erythroid tissues may be due to Kell glycoprotein may also be involved in the interaction of
increased intracellular proteolytic degradation of XK. In endothelial cells with tumor cells but we did not detect Kell
transfected cells expressing both Kell glycoprotein and XK, expression in endothelial cell lines (HUVEC) or HUVSM by
the Kell/XK complex is formed early in the endoplasmic retic- RT-PCR (data not shown), although, according to the Tran-
ulum during transport to the cell surface, and although Kell scription profiling of human cell lines and tissues of the EBI
glycoprotein is not absolutely required for XK transport to the data base, Kell transcripts were detected in CD105 (Endo-
cell surface, in Kel(–/–) cells the lack of Kell glycoprotein may glin) positive endothelial cells (samples 32 and 106, Array
hamper XK transport, resulting in XK being more readily Express Warehouse 8.10). CD105 is a proliferation-associ-
degraded [41]. Thus, XK may be more susceptible to degra- ated and hypoxia-inducible protein abundantly expressed in
dation as a free molecule than it is in its normal form as part angiogenic endothelial cells [42].
of the disulfide bonded Kell/XK complex. We did not detect severe motor function pathology in
The expression of Kell glycoprotein in humans was origi- mice lacking Kell. However, some motor characteristics
nally thought to be limited to erythroid tissues, but later could be identified as part of the phenotype of mice lacking
reports suggested that Kell glycoprotein is also expressed the Kell gene. Mice lacking Kell glycoprotein demonstrated
in testis and in other nonerythroid tissues including brain wider hind limb landing foot splay in the drop test, and,
and skeletal muscle [1,30,32]. In mice, analysis of expres- although statistical significance was not reached, in rotarod
sion of Kell glycoprotein by in situ hybridization revealed testing generally had more falls during training and a
that Kell glycoprotein expression is limited to erythroid tis- shorter time to fall during testing at increasing speeds.
sues with a small amount in testis [31]. The earlier results In conclusion, the phenotypes in Kell knockout mice sug-
indicating the presence of Kell glycoprotein in nonerythroid gested possible physiological functions of Kell glycoprotein
tissues in humans may have been partly due to the pres- in heart, red cell ion transport, neovascularization in
ence of residual blood in the tissues examined. Our current tumors, and motor function. The targeted disruption of Kel
RT-PCR studies indicate that Kell transcript is present in in mouse enabled us to identify phenotypes that would not

496 American Journal of Hematology


be easily detected in humans lacking Kell glycoprotein. In
this regard, the Kell knockout mouse provides a good
animal model for the study of normal and/or pathophysio-
logical functions of Kell.
Materials and Methods
Targeting vector. Targeted disruption of Kel was accomplished by ho-
mologous recombination using a targeting vector constructed in pBlue-
script using a Kel clone that has a 12.9 kb EcoRI-EcoRI fragment which
contains exons 11 to 19. The Kel clone was selected by screening the
Mouse 129 SvJ genomic library (Stratagene, La Jolla, CA). A 3.2 kb seg-
ment of Kel, spanning a region from exon 11 to part of exon 16 (HindIII to
EcoRV fragment) and which includes the region encoding the zinc bind-
ing, enzymatically active site of Kell glycoprotein, was replaced with a 1.6
kb cassette containing the neomycin-resistance gene (Supporting Infor-
mation Fig. 1). A 2.7 kb segment of intron 10 was placed in the 5’ short
arm of the construct and a 7 kb fragment downstream of intron 16 was
placed in the 3’ long arm. The EcoRV site at the 5’ end of the long arm
was destroyed and a BamHI site at the 3’ end of short arm was created
when the neomycin gene was inserted, resulting in the generation of a 15
kb band when cut by EcoRV and a 3.7 kb band when cut by BamHI,
compared to 8 and 9 kb bands, respectively, in the wild type allele.
Generation of mice carrying the targeted allele. RW-4 ES cells
derived from the 129/SvJ mouse strain were electroporated with the lin-
earized targeting vector and the positive cells were selected using
G418 in the culture medium. Homologous recombinants were identified
by Southern blot analysis of EcoRV or BamHI cut genomic DNA. Chi-
meras were produced using morula aggregation technology [43]
(GenomeSystems, Inc). The germ line transmitted heterozygous Kel
knockout [Kel(–/–)] mice were selected by mating chimeras (male and
female) with C57BL/6 mice. The genetic background of the knockout
mice was made fully congenic by back-crossing the hybrids with
C57BL/6 for at least 10 generations. Finally, homozygous Kel(–/–) mice
were generated by mating the Kel(1/–) mice.
Screening of Knockout mice by Southern blot and genomic DNA
PCR, and characterization by Western blot.
Southern blot. DNA obtained from mouse tails was cut with BamHI
or EcoRV followed by agarose gel electrophoresis. The separated DNA
was blotted on Zeta-Probe GT Blotting Membrane (Bio-Rad). BamHI
cut DNA was hybridized with a 600bp 5’ end probe that distinguishes
the wild type Kel gene (9 kb band) from the knockout allele, Kel(–/),
(3.7 kb band). EcoRV cut DNA was hybridized with a 700 bp 3’ end
probe that gives an 8 kb band for the wild type allele and a 15 kb band
for Kel(–/) allele. The size and the location of the probes are described
in the construct map (Supporting information Fig. 1).
PCR. The following primers were used for screening the tail genomic
DNA by PCR.
Kel WT:(Min10FA) 5’-GCCTGAGTTGTTTTCCTATCCCTTGG-3’
(Min11RA) 5’-GCACACCACAAAGAAGTATACACTTC-3’
The resulting amplicon is 349 bp in size
Kel KO: (3’probF) 5’-AGGAGTCAGTGGCAGTACCAGATTCTAG-3’
(NeoS) 5’-GGATGCGGTGGGCTCTATGGCTTCTG-3’
The resulting amplicon is 2.8 kb in size. This primer set was used to
confirm the integrity of the targeted gene at the 5’ end recombination
site. Subsequently, the following primer set, which amplifies a shorter
segment with an improved signal was used for identification of the
Kel(–/) allele.
mKel-WTKO-1F: 5’-GTGCCTGAGTTGTTTTCCTATCCCTTGG-3’
mKel-KO-1R: 5’-CATTCGACCACCAAGCGAAACATCGCAT-3’
The resulting amplicon is 941 bp in size.
Western blot. Blood was collected into EDTA-containing tubes and
red cell ghosts were prepared with low salt buffer (5 mM sodium phos-
phate buffer, pH 7.4 containing 10 U/mL trasylol). The ghosts were dis-
solved in 23 sample buffer with or without 1 mM DTT and the sample
was separated in 10% SDS-PAGE. Kell glycoprotein was probed with a
rabbit antibody to a peptide of mouse Kell (mKell 358-393), and XK
was probed with rabbit antibody to a mouse XK peptide (mXK 404-
443) using a western blot kit (SupersignalTM, Pierce).
Determination of endothelin converting activity of Kell glycoprotein.
Endothelin converting activities of RBCs of Kel(–/–), Kel(1/–), and WT
mice were analyzed using the ET-1 enzyme immunoassay (EIA) Kit
(Cayman Chemical, Ann Arbor, MI) following the company protocol as
described previously [8].
RT-PCR. The RT-PCR was carried out as described previously [31].
Physiologic examination of heart function. Electrocardiogram and
blood pressure. Six conscious 1-year-old female mice underwent a
American Journal of Hematology
research article
single continuous ECG test at the Jackson Laboratory (West Sacra-
mento, CA) using the noninvasive AnonyMOUSE ECG screening sys-
tem (Mouse Specifics, Inc). Systolic blood pressure testing was also
conducted over 30 consecutive cycles of cuff inflation and deflation (10
preliminary and 20 test measurements) using the Visitech BP-2000
blood pressure analysis system.
Hematological analysis. Blood samples (25 lL) from 25 male WT
and 17 male Kel(–/–) mice of 153 to 173 days of age were collected by
retro-orbital sinus bleeding using heparinized capillary tubes. Blood
was mixed with 100 lL RPMI and further diluted with 100 lL of 1%
BSA in RPMI before analysis. The tests were performed in two groups
on two different days. Group one contained 18 WT and 9 Kel(–/–) mice
and group two contained 7 WT and 8 Kel(–/–) mice. Complete blood
analysis was performed using the Advia 120 Hematology System
(Bayer, Tarrytown, NY).
Measurement of Gardos channel activity of RBCs and the effect of
ET-3.
Erythrocyte preparation. The procedure for the preparation of eryth-
rocytes was followed as described previously [39].
Measurement of 86Rb influx. The measurement of 86Rb influx was
carried out by a previously described method [39]. Briefly, the influx
medium contained 1 mM ouabain (Sigma chemical, St. Luis, MO) 10
mM Tris-MOPS, pH 7.4 (228C), 10 lM bumetanide, and 10 lCi/ml
86Rb with and without 700 nM ET-3 (Sigma chemicals) and 50 nM
charybdotoxin (ChTX, Sigma chemical). Preincubations with ET-3 were
carried out for 15 min at 378C in an isotonic Na-media. Free ionic Ca21
in the influx media was buffered to 7 lM with 1 mM citrate buffer. Ca21
concentration was calculated using the dissociation constants for citrate
and correcting for ionic strength in the presence of 0.15 mM MgCl2 at
pH 7.4. At time 0 min, A23187 ionophore (5 lM) (Calbiochem-Novabio-
chem Corp., La Jolla, CA) was added and aliquots were removed at 2
and 5 min and the erythrocyte-associated radioactivity was counted as
described previously [40].
Tumor growth and neovascularization. The highly tumorigenic and
weakly metastatic Lewis lung carcinoma cells (LL/2 cell line originally
derived from a C57BL/6 mouse, ATCC) were injected subcutaneously
on the backs of Kel(–/–) and WT mice (106 cells/ mouse). Before
injection, the possible presence of adventitious and endogenous
viruses in the cultured LL/2 cells was assessed by MAP (Mouse Anti-
body Production) test at the Diagnostic laboratory of Charles River Lab-
oratories. By 2 or 3 weeks following injection, the tumors reached 1
cm3, at which point the mice were euthanized and the shape, weight,
and vascularization of the tumors were measured. The formation of
vascular structures in and around the tumor was examined by light mi-
croscopy. Two sets of experiment were performed, the first using 12-
month-old mice and the second using 5-month-old mice. Tumor tissues
were fixed in IHC Zinc Fixative (Formalin-Free) solution (BD Bioscien-
ces, San Jose, CA) for not more than 24 hr, embedded in paraffin
blocks and sectioned (5–6 lm thickness).
Immunohistochemical staining of tissue sections for CD31 (PECAM-1).
Tissue sections were examined for neovascularization by immunostain-
ing with anti-CD31 (BD biosciences) using the TSATM Biotin System
(Perkin Elmer, Waltham, MA) in conjunction with NovaREDTM (Vecter
Labs, Orton Southgate, Peterborough, UK) visualization reagent follow-
ing the manufacturer’s protocol. Antigenic epitopes were retrieved by
digestion with Trypsin (Sigma, St. Louis, MO) at 378C, for 10 min. Endog-
enous peroxidase was blocked with 3% H2O2 for 5 min. The area selec-
tion for counting the microvessels was performed, first, by viewing the
tumor with a low power objective lens (43) to find the center part of the
tumors and, subsequently, by randomly selecting 1–3 fields depending
on the tumor size with a higher power objective lens (203) . Counting of
microvessels was performed at 2003 magnification using SPOT 4.1
software.The per-spot values from each tumor were averaged. The aver-
aged value was used in the statistical analysis to represent one tumor.
Evaluation of motor function. Design and testing schedule. Six WT
and 6 Kel(–/–) male mice were assessed at 6 and 12 months of age
with grip strength, ataxia, wire suspension, and spontaneous motor
activity tests. Additionally, an observational assessment, the Functional
Observational Battery, was conducted monthly. At 18 months of age,
mice were added to the groups [n 5 17 Kel (1/1), n 5 15 Kel(–/–)] for
a more challenging motor test, the incremental speed rotarod. Tests
were performed at the Murine Behavioral Assessment Laboratory of
the University of California-Davis using standardized protocols and
have been described previously [44–47]. Briefly, the Functional Obser-
vational Battery is a 22-item neurological screen for abnormal reflexes,
motor patterns, seizure activity, and autonomic signs. For the ataxia
test, stride length is measured as the distance between successive
497
research article
paw placements as the mouse walks down a narrow path, and landing
foot splay in the drop test is recorded as the distance between paws
when the mouse is dropped from a height of 16 cm. Grip strength is
measured as the force required to break the grip reflex. For wire sus-
pension, the mouse grasps a 2 mm wire and has 60 sec to make its
way to a side support. The spontaneous activity tests used beam break
technology to track movement in smaller (10 x 10 inch) and larger (16 x
16 inch) arenas for shorter (90 min) or longer (20 hr) evaluation periods.
Rotarod testing included 3 days of training at 24 rpm followed by two
60-sec trials at speeds of 8, 16, 24, 30, 33, and 44 rpm [48].
Statistical analysis. All statistical analysis except data for motor func-
tion evaluation was performed by Student’s t-test (unpaired two-tail anal-
ysis) or two-way ANOVA using GraphPad Prism software (La Jolla, CA).
For the motor function evaluation data, a single factor ANOVA was
conducted with Kel genotype as the independent variable. Repeated
measures ANOVA (RMANOVA) was used for the sequential tests. If
homogeneity of variance across groups was not observed, the Welch
ANOVA, which does not assume equal variance, was used. All statistical
analyses were performed with JMP software (SAS, Cary, NC).
Acknowledgments
The authors thank the staff of the Nucleic Acid Analysis
Laboratory of the New York Blood Center for DNA sequenc-
ing. They also thank Dr. Eric C. B. Milner of BioMedical
Science Writers, LLC, for helping in editing the manuscript.
References
1. Lee S, Zambas ED, Marsh WL, et al. Molecular cloning and primary struc-
ture of Kell blood group protein. Proc Natl Acad Sci USA 1991;88:6353–
6357.
2. Lee S. The value of DNA analysis for antigens of the Kell and Kx blood group
systems. Transfusion 2007;47:32S–39S.
3. Lee S, Russo DC, Reiner AP, et al. Molecular defects underlying the Kell null
phenotype. J Biol Chem 2001;276:27281–27289.
4. Carbonnet F, Hattab C, Collec E, et al. Immunochemical analysis of the Kx
protein from human red cells of different Kell phenotypes using antibodies
raised against synthetic peptides. Br J Haematol 1997;96:857–863.
5. Ho M, Chelly J, Carter N, et al. Isolation of the gene for McLeod syndrome that
encodes a novel membrane transport protein. Cell 1994;77:869–880.
6. Russo D, Redman C, Lee S. Association of XK and Kell blood group proteins.
J Biol Chem 1998;273:13950–13956.
7. Lee S, Russo D, Redman CM. The Kell blood group system: Kell and XK
membrane proteins. Semin Hematol 2000;37:113–121.
8. Lee S, Lin M, Mele A, et al. Proteolytic processing of big endothelin-3 by the
Kell blood group protein. Blood 1999;94:1440–1450.
9. Sha Q, Redman CM, Lee S. Endothelin-3-converting enzyme activity of the
KEL1 and KEL6 phenotypes of the Kell blood group system. J Biol Chem
2006;281:7180–7182.
10. Claperon A, Hattab C, Armand V, et al. The Kell and XK proteins of the Kell
blood group are not co-expressed in the central nervous system. Brain Res
2007;1147:12–24.
11. Garcia-Barcelo M, Sham MH, Lee WS, et al. Highly recurrent RET mutations
and novel mutations in genes of the receptor tyrosine kinase and endothelin
receptor B pathways in Chinese patients with sporadic Hirschsprung disease.
Clin Chem 2004;50:93–100.
12. Crawley DE, Liu SF, Barnes PJ, et al. Endothelin-3 is a potent pulmonary va-
sodilator in the rat. J Appl Physiol 1992;72:1425–1431.
13. Feletou M, Tang EH, Vanhoutte PM. Nitric oxide the gatekeeper of endothelial
vasomotor control. Front Biosci 2008;13:4198–4217.
14. Takasuka T, Adachi M, Miyamoto C, et al. Characterization of endothelin recep-
tors ETA and ETB expressed in COS cells. J Biochem 1992;112:396–400.
15. Anita I, Yaira M, Maria Del RG. Endothelin signaling pathways in rat adrenal
medulla. Cell Mol Neurobiol 2006;26:703–718.
16. Leite-Moreira AF, Bras-Silva C. Inotropic effects of ETB receptor stimulation
and their modulation by endocardial endothelium, NO, and prostaglandins.
Am J Physiol Heart Circ Physiol 2004;287:H1194–H1199.
17. Nagy N, Goldstein AM. Endothelin-3 regulates neural crest cell proliferation
and differentiation in the hindgut enteric nervous system. Dev Biol 2006;293:
203–217.
18. Spinella F, Rosano L, Di CV, et al. Endothelin-1 and endothelin-3 promote
invasive behavior via hypoxia-inducible factor-1alpha in human melanoma
cells. Cancer Res 2007;67:1725–1734.
498
19. Ziche M, Morbidelli L, Donnini S, et al. ETB receptors promote proliferation
and migration of endothelial cells. J Cardiovasc Pharmacol 1995;26(Suppl
3):S284–S286.
20. Goligorsky MS, Budzikowski AS, Tsukahara H, et al. Co-operation between
endothelin and nitric oxide in promoting endothelial cell migration and angio-
genesis. Clin Exp Pharmacol Physiol 1999;26:269–271.
21. Makita T, Sucov HM, Gariepy CE, et al. Endothelins are vascular-derived axo-
nal guidance cues for developing sympathetic neurons. Nature 2008;452:
759–763.
22. Nelson J, Bagnato A, Battistini B, et al. The endothelin axis: Emerging role in
cancer. Nat Rev Cancer 2003;3:110–116.
23. Bagnato A, Spinella F, Rosano L. The endothelin axis in cancer: The promise
and the challenges of molecularly targeted therapy. Can J Physiol Pharmacol
2008;86:473–484.
24. Wu WS. The signaling mechanism of ROS in tumor progression. Cancer
Metastasis Rev 2006;25:695–705.
25. Barton M, Yanagisawa M. Endothelin: 20 years from discovery to therapy.
Can J Physiol Pharmacol 2008;86:485–498.
26. Yanagisawa H, Yanagisawa M, Kapur RP, et al. Dual genetic pathways of
endothelin-mediated intercellular signaling revealed by targeted disruption of
endothelin converting enzyme-1 gene. Development 1998;125:825–836.
27. Hosoda K, Hammer RE, Richardson JA, et al. Targeted and natural (piebald-
lethal) mutations of endothelin-B receptor gene produce megacolon associ-
ated with spotted coat color in mice. Cell 1994;79:1267–1276.
28. Baynash AG, Hosoda K, Giaid A, et al. Interaction of endothelin-3 with endo-
thelin-B receptor is essential for development of epidermal melanocytes and
enteric neurons. Cell 1994;79:1277–1285.
29. Puffenberger EG, Hosoda K, Washington SS, et al. A missense mutation of
the endothelin-B receptor gene in multigenic Hirschsprung’s disease. Cell
1994;79:1257–1266.
30. Camara-Clayette V, Rahuel C, Lopez C, et al. Transcriptional regulation of the
KEL gene and Kell protein expression in erythroid and non-erythroid cells.
Biochem J 2001;356:171–180.
31. Lee S, Sha Q, Wu X, et al. Expression profiles of mouse Kell, XK, and
XPLAC mRNA. J Histochem Cytochem 2007; 55:365–374.
32. Russo D, Wu X, Redman CM, et al. Expression of Kell blood group protein in
nonerythroid tissues. Blood 2000;96:340–346.
33. Lee S, Russo DC, Pu J, et al. The mouse Kell blood group gene (Kel): cDNA
sequence, genomic organization, expression, and enzymatic function. Immu-
nogenetics 2000;52:53–62.
34. Shmueli O, Horn-Saban S, Chalifa-Caspi V, et al. GeneNote: Whole genome
expression profiles in normal human tissues. C R Biol 2003;326:1067–1072.
35. Chalifa-Caspi V, Shmueli O, Jamin-Rodrig H, et al. GeneAnnot: Interfacing
GeneCards with high-throughput gene expression compendia. Brief Bioinform
2003;4:349–360.
36. Safran M, Chalifa-Caspi V, Shmueli O, et al. Human Gene-Centric Databases
at the Weizmann Institute of Science: GeneCards, UDB, CroW 21 and
HORDE. Nucl Acids Res 2003;31:142–146.
37. Danson EJ, Paterson DJ. Cardiac neurobiology of nitric oxide synthases. Ann
N Y Acad Sci 2005;1047:183–196.
38. Herring N, Danson EJ, Paterson DJ. Cholinergic control of heart rate by nitric
oxide is site specific. News Physiol Sci 2002;17:202–206.
39. Rivera A, Rotter MA, Brugnara C. Endothelins activate Ca(21)-gated K(1)
channels via endothelin B receptors in CD-1 mouse erythrocytes. Am J Phys-
iol 1999;277:C746–C754.
40. Rivera A, Jarolim P, Brugnara C. Modulation of Gardos channel activity by
cytokines in sickle erythrocytes. Blood 2002;99:357–603.
41. Russo D, Lee S, Redman C. Intracellular assembly of Kell and XK blood
group proteins. Biochim Biophys Acta 1999;1461:10–18.
42. Fonsatti E, Altomonte M, Nicotra MR, et al. Endoglin (CD105): A powerful
therapeutic target on tumor-associated angiogenetic blood vessels. Oncogene
2003;22:6557–6563.
43. Wood SA, Pascoe WS, Schmidt C, et al. Simple and efficient production of
embryonic stem cell-embryo chimeras by coculture. Proc Natl Acad Sci USA
1993;90:4582–4585.
44. Golub MS, Germann SL, Mercer M, et al. Behavioral consequences of ovar-
ian atrophy and estrogen replacement in the APPswe mouse. Neurobiol
Aging 2008;29:1512–1523.
45. Jaubert PJ, Golub MS, Lo YY, et al. Complex, multimodal behavioral profile of
the Homer1 knockout mouse. Genes Brain Behav 2007;6:141–154.
46. Golub MS, Germann SL, Araiza RS, et al. Movement disorders in the Hfe
knockout mouse. Nutr Neurosci 2005;8:239–244.
47. Golub MS, Germann SL, Lloyd KC. Behavioral characteristics of a nervous
system-specific erbB4 knock-out mouse. Behav Brain Res 2004;153:159–
170.
48. Tomemori Y, Ichiba M, Kusumoto A, et al. A gene-targeted mouse model for
chorea-acanthocytosis. J Neurochem 2005;92:759–766.
American Journal of Hematology