Báo cáo khóa học: Determination of modulation of ligand properties of synthetic complex-type biantennary N-glycans by introduction of bisecting GlcNAc in silico, in vitro and in vivo pot

Computational analysis of glycan conformation with pro-longed simulation periods in vacuo and in a solvent box revealed two main effects: backfolding of the a1–6 arm and stacking of the b

Determination of modulation of ligand properties of synthetic

complex-type biantennary N-glycans by introduction of bisecting GlcNAc in silico , in vitro and in vivo

Sabine Andre´1, Carlo Unverzagt2,*, Shuji Kojima3, Martin Frank4, Joachim Seifert2,†, Christian Fink5, Klaus Kayser6, Claus-Wilhelm von der Lieth4and Hans-Joachim Gabius1

1

Institut fu¨r Physiologische Chemie, Tiera¨rztliche Fakulta¨t, Ludwig-Maximilians-Universita¨t Mu¨nchen, Germany;2Institut fu¨r Organische Chemie und Biochemie, Technische Universita¨t Mu¨nchen, Garching, Germany;3Faculty of Pharmaceutical Sciences, Tokyo University of Science, Japan;4Zentrale Spektroskopie, Deutsches Krebsforschungszentrum, Heidelberg, Germany;

5 Radiologische Diagnostik und Therapie, Deutsches Krebsforschungszentrum, Heidelberg, Germany; 6 Institut fu¨r Pathologie, Charite´, Humboldt Universita¨t, Berlin, Germany

We have investigated the consequences of introducing

a bisecting GlcNAc moiety into biantennary N-glycans

Computational analysis of glycan conformation with

pro-longed simulation periods in vacuo and in a solvent box

revealed two main effects: backfolding of the a1–6 arm and

stacking of the bisecting GlcNAc and the neighboring Man/

GlcNAc residues of both antennae.Chemoenzymatic

syn-thesis produced the bisecting biantennary decasaccharide

N-glycan and its a2–3(6)-sialylated variants.They were

conjugated to BSA to probe the ligand properties of

N-glycans with bisecting GlcNAc.To assess affinity

altera-tions in glycan binding to receptors, testing was performed

with purified lectins, cultured cells, tissue sections and

ani-mals.The panel of lectins, including an

adhesion/growth-regulatory galectin, revealed up to a sixfold difference in

affinity constants for these neoglycoproteins relative to data

on the unsubstituted glycans reported previously [Andre´, S ,

Unverzagt, C , Kojima, S , Dong, X , Fink, C , Kayser, K &

Gabius, H.-J (1997) Bioconjugate Chem 8, 845–855].The enhanced affinity for galectin-1 is in accord with the increased percentage of cell positivity in cytofluorimetric and histochemical analysis of carbohydrate-dependent binding

of labeled neoglycoproteins to cultured tumor cells and routinely processed lung cancer sections.Intravenous injec-tion of iodinated neoglycoproteins carrying galactose-ter-minated N-glycans into mice revealed the highest uptake in liver and spleen for the bisecting compound compared with the unsubstituted or core-fucosylated N-glycans.Thus, this substitution modulates ligand properties in interactions with lectins, a key finding of this report.Synthetic glycan tailoring provides a versatile approach to the preparation of newly substituted glycans with favorable ligand properties for medical applications

Keywords: bisecting GlcNAc; drug targeting; lectin; neo-glycoprotein; tumor imaging

A major challenge in the post-genomic era is the

functional analysis of post-translational protein

modifica-tions leading to medical applicamodifica-tions.With about two

thirds of eukaryotic protein sequences reported to harbor the N-glycosylation sequon, this type of modification is typical of membrane and secretory proteins [1].The complex enzymatic machinery located in the endoplasmic reticulum and Golgi network, representing a notable investment in terms of genomic coding capacity, is known

to produce a large variety of N-glycans [2,3].These two aspects, i.e frequent protein glycosylation and large panel

of glycosyltransferases, suggest a nonrandom, albeit not template-driven, synthesis with a strong impact on cellular activities [3,4].The glycan chains produced, referred to as the glycomic profile, reflect cellular parameters such as differentiation and disease processes [3,5].Current efforts are directed to relating distinct glycan sequence motifs to key effector mechanisms at the cellular level.In this context our study focuses on defining the role of the bisecting GlcNAc residue

A key regulatory step in N-glycan processing is the addition of a b1–4-linked GlcNAc residue to the central b-mannose unit of the core pentasaccharide.This reaction

is catalyzed by N-acetylglucosaminyltransferase III (EC 2.4.1.144; GnT-III) placing this GlcNAc residue (Fig.1, bottom panel) between the a1–3 and a1–6 arms,

Correspondence to S.Andre´, Institut fu¨r Physiologische Chemie,

Tiera¨rztliche Fakulta¨t, Ludwig-Maximilians-Universita¨t Mu¨nchen,

Veterina¨rstr.13, 80539 Mu¨nchen, Germany.

Fax: + 49 89 2180 2508, Tel.: + 49 89 2180 2290,

E-mail: Sabine.Andre@lmu.de

Abbreviations: E-PHA, erythroagglutinating phytohemagglutinin;

GnT-III, N-acetylglucosaminyltransferase III; MD, molecular

dynamics.

Enzymes: N-acetylglucosaminyltransferase III (EC 2.4.1.144;

GnT-III).

*Present address: Bioorganische Chemie, Geba¨ude NW1, Universita¨t

Bayreuth, 95440 Bayreuth, Germany.

Present address: 3 Hi-Tech Court, Brisbane Technology Park,

Eight Mile Plains, Brisbane, QLD 4113, Australia.

Dedication: dedicated to and in thankful commemoration of Professor

F.Cramer who died recently three months before his 80th birthday.

(Received 25 September 2003, revised 28 October 2003,

accepted 6 November 2003)

which extend from the core pentasaccharide [2].The

introduction of a bisecting GlcNAc residue will probably

perturb conformational aspects proximal and also distal to

its position in the glycan chain.Evidence has been provided

for a shift of the conformational equilibrium of the torsional

angle x of the C5–C6 bond of the a1–6 linkage between the

two positions around 180 and)60  to the 180  position

for the biantennary complex-type N-glycan and to the)60 

position for the hybrid-type N-glycan [6–11].Next, in the

presence of chain elongation beyond the core

pentasaccha-ride (but not in its absence), the a1–3 linkage becomes

considerably more restrained than in the N-glycan without a

bisecting GlcNAc unit [12–14].In summary, these reports of

NMR experiments and molecular modeling intimate that

the bisecting GlcNAc acts like a wedge with implications for

the shape of the N-glycan.The occurrence of distinct

alterations in shape induced by this substitution prompted

us to examine the ligand properties of the substituted

N-glycans in detail.Undoubtedly, this work would gain

substantial relevance and importance, if biological effects

were known associated with cellular expression of this

substitution

Indeed, the second premise for our study was provided

by respective reports.The applied methods included up-regulation of GnT-III activity, studies on normal or tumor cells with natural or increased GnT-III activity, ectopic GnT-III expression, and generation of GnT-III-deficient mice.In detail, the mutant phenotype of LEC10 CHO cells displaying resistance to ricin was attributed to induction of GnT-III activity [15,16].Transcriptional up-regulation of GnT-III gene expression by forskolin and thus increased production of N-glycans with bisecting GlcNAc led to the decreased cell surface presence of lysosomal-associated membrane glycoprotein-1 and c-glutamyltrans-peptidase [17].This exemplary result links the bisecting motif to intracellular routing.Regarding tumors, GnT-III was markedly increased in the preneoplastic stage of rat (but not mouse) liver carcinogenesis, the blast crisis of chronic myelogenous leukemia, and pediatric brain tumors [18].The bisecting GlcNAc then makes its presence felt in tumor progression.This process can apparently be influenced nonuniformly, as attested by analysis of different cell types GnT-III overexpression yielded suppression of lung meta-stasis for mouse B16 melanoma cells, increased adhesivity

Fig 1 Neoglycoproteins with BSA as carrier

for the test panel of six complex-type

bianten-nary N-glycans The upper part shows the

structures of the biantennary nonasaccharide

(Bi9) and its a2–6(3)-sialylated

undecasac-charides (Bi1126, Bi1123).The bottom panel

illustrates the corresponding decasaccharides

and dodecasaccharides with bisecting

Glc-NAc.For expample, the abbreviation

BiB1226-BSA stands for biantennary bisecting

N-glycan 12-mer a2–6-sialylated BSA

conju-gate.The linker arm and the attachment point

to an e-amino group of lysine of the carrier

protein are also shown.

and decreased migration activity was observed for human

U373 glioma cells bearing mutual overexpression of

GnT-III relative to GnT-V (shift from biantennary N-glycans

with bisecting GlcNAc to highly branched N-glycans), and

GnT-III-overexpressing human K562 erythroleukemic cells

rather effectively colonized the spleen in athymic mice with

acquisition of resistance to NK cell cytotoxicity [19–21]

Moreover, hemopoiesis supported by bone marrow stroma

requires, in part, integrity of N-glycans, and hemopoietic

dysfunction with lowered production of nonadherent cells in

transgenic mice with GnT-III overexpression demonstrates

that respective changes in the N-glycan profile account for

suppression of proliferation and differentiation of

hemopoi-etic cells [22,23].Altered glycosylation appears to translate

into cellular responses via biochemical pathways.Apart

from having a bearing on the protein part’s folding or access

to binding partners, the capacity of protein-bound glycan

chains to impart a discrete recognitional role to the protein

[24] deserves attention in this respect.Explicitly, the bisecting

GlcNAc has potential either to be directly engaged in

intermolecular interactions with receptors, e.g endogenous

lectins [4,25–30], or to modulate binding processes at other

sites in the chain through the shape alterations induced by its

presence.That bisecting GlcNAc constitutes a docking point

for protein–carbohydrate interactions is known from the

plant lectin erythroagglutinating phytohemagglutinin

(E-PHA), whereas the presence of a bisecting GlcNAc in a

core-fucosylated biantennary N-glycan did not affect

bind-ing to the L-fucose-specific lectin from Aleuria aurantia

[31–35].Annexin V, a cell surface marker of apoptotic cells,

has recently been reported to share this property with the

plant lectin E-PHA [36]

To determine if and to what extent ligand properties are

conveyed and/or affected by this natural substitution, and

then to devise routes for potential medical application, we

combined synthetic, biochemical, cell biological,

histopatho-logical and in vivo procedures.The question of whether

introduction of the bisecting GlcNAc changes the ligand

properties (binding affinities) of biantennary N-glycans for

branch-end-specific lectins as well as for cells and organs has

so far not been tackled.To this end, it is desirable to obtain

pure test substances, i.e N-glycans free of natural

micro-heterogeneity, by a convenient preparative route.After

developing a chemoenzymatic synthesis for biantennary

N-glycans with this distinct substitution, the preparation of

the corresponding glycan-bearing neoglycoproteins made

the required tools available.In a previous study, we prepared

neoglycoproteins carrying unsubstituted biantennary

N-glycans [37].These compounds were examined for their

properties in enzyme-linked lectin and cell binding assays,

histopathological monitoring, and organ biodistribution

[37].The utility of our combined approach and the

validity of the concept were then tested by assessing the

properties of N-glycans modified with a1–6 core fucose

Changes in cell biological parameters could be attributed

to the N-glycan modification [38].Pursuing this line of

research, we report results obtained with neoglycoproteins

carrying biantennary N-glycans with the bisecting

Glc-NAc motif.The chemoenzymatic synthesis of bisecting

N-glycans with terminal galactose or sialic acid residues is

presented followed by the generation of neoglycoproteins

To take advantage of the enormous strides made over

recent years in acquiring, storing and handling increasing quantities of data from molecular modeling calculations,

we devoted part of the study to scrutinizing the conformational properties of the substituted and unsub-stituted N-glycans.Their behavior patterns were compar-atively analyzed by refined computational protocols

in vacuo and, compared with previous calculations, in the presence of a defined solvent.The simulation periods were significantly extended.Experimentally, the ligand characteristics of the N-glycans as part of the neoglyco-proteins (Fig.1) were evaluated in biochemical, cell biological and histopathological assays including deter-mination of biodistribution in vivo relative to the control compounds (complex-type biantennary N-glycans lacking bisecting GlcNAc).It is shown that the bisecting substi-tution triggers alteration of ligand properties, implying functional significance of the bisecting GlcNAc as a modulator of N-glycan biorecognition

Materials and methods

Synthetic and analytical procedures NMR spectra were recorded with a Bruker AMX 500 spectrometer.HPLC separations were performed on a Pharmacia LKB gradient system 2249 equipped with a Pharmacia LKB Detector VWM 2141 (Freiburg, Germany) For size-exclusion chromatography, a Pharmacia Hi Load Superdex 30 column (600· 16 mm) was used.RP-HPLC was performed on a Macherey-Nagel Nucleogel RP 100-10 column (300· 25 mm).BSA, b1–4-galactosyltransferase, a2–6-sialyltransferase and nucleotide sugars were purchased from Sigma (Munich, Germany), and alkaline phosphatase (calf intestine, molecular biology grade; EC 3.1.3.1) from Roche Diagnostics (Heidelberg, Germany).We are grateful

to J.C.Paulson (Cytel Corp., San Diego, USA) for a sample

of recombinant a2–3-sialyltransferase.ESI mass spectra were recorded on a Finnigan TSQ 700 in methanol/water MALDI-TOF mass spectra were collected by D.Renauer at the Boehringer Mannheim research facility (Tutzing, Ger-many) on a Voyager Biospectrometry workstation (Vestec/ Perseptive) MALDI-TOF mass spectrometer, using 2,5-dihydroxybenzoic acid as a matrix.The structures of the synthetic N-glycans were invariably confirmed by the following 2D NMR experiments: TOCSY, NOESY, HMQC (heteronuclear multiple-quantum coherence), HMQC-COSY, HMQC-DEPT (distortionless enhance-ment by polarization transfer), HMQC-TOCSY.Signals of NMR spectra were assigned according to the following convention including the spacer

N1 -(6-Benzyloxycarbonyl-6-aminohexanoylamido)-2-ace-tamido-2-deoxy-b-D-glucopyranosyl)-(1fi 2)-a-D -mann-opyranosyl-(1fi 3)-[2-acetamido-2-deoxy-b-D -glucopy-ranosyl-(1fi 4)]-[2-acetamido-2-deoxy-b-

-glucopyran-osyl-(1fi 2)-a-D-mannopyranosyl-(1fi 6)]-b-D

-mann-opyranosyl-(1fi

4)-2-acetamido-3,6-di-O-benzyl-2-de-oxy-b-D-glucopyranosyl-(1fi

4)-2-acetamido-3,6-di-O-benzyl-2-deoxy-b-D-glucopyranoside (3;

Bzl3-BiB8AH-Z).To a portion of 61.1 mg (32.1 lmol) glycosylazide (1)

dissolved in 2 mL methanol were added 65 lL

triethylam-ine.The flask was flushed with argon followed by addition

of 200 lL propane-1,3-dithiol [39].After completion of the

reaction (3.5 h; TLC: Rf amine¼ 0.22; propan-2-ol/1M

ammonium acetate, 4 : 1, v/v), the solution was evaporated

and dried in high vacuo for 15 min.The remainder was

allowed to react with a solution of activated

Z-aminohex-anoic acid (2) prepared as follows: 236 mg (0.89 mmol,

20 eq.) Z-aminohexanoic acid were dissolved in 3.5 mL

N-methylpyrrolidone Subsequently, 136.3 mg (0.89 mmol,

20 eq.) N-hydroxybenzotriazol (HOBT), 285.8 mg

(0.89 mmol, 20 eq.) TBTU

[(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium tetrafluoroborate] and 233 lL

(1.32 mmol) di-isopropylethylamine were added, following

a protocol for standard peptide chemistry activation [40]

The mixture was stirred until a clear solution was obtained

and adjusted to pH¼ 9 by adding 130 lL

di-isopropyl-ethylamine

The dried glycosylamine was dissolved in 2.5 mL of the

solution prepared as above, and the pH was adjusted to

9.0 with di-isopropylethylamine After 30 min at ambient

temperature (TLC: propan-2-ol/1M ammonium acetate,

4 : 1, v/v), 0.5 mL of the solution was added and stirring

was continued for 5 min.The reaction mixture was

evaporated and dried in high vacuo.Purification of the

remainder was performed by RP-HPLC [acetonitrile/water,

gradient of 35–45% acetonitrile run in 40 min; flow

rate¼ 8 mLÆmin)1, Macherey–Nagel Nucleogel RP

100-10 (300· 25 mm)] The yield was 36.8 mg (54.0%), and

analytical data were as follows: Rf(amine)¼ 0.22

(propan-2-ol/1M ammonium acetate, 2 : 1) and Rf (3)¼ 0.47

(propan-2-ol/1Mammonium acetate, 2 : 1);½a22D ¼) 2.3

(0.7, CH3CN/H2O, 1 : 1) and C100H139N7O43

(M¼ 2127.22) ESI-MS: Mcalcd¼ 2125.9, Mfound¼

1064.3 (M + 2 H)2+.(Complete set of1H/13C-NMR data

not shown.)

N1-(6-Aminohexanoylamido)-2-acetamido-2-deoxy-b-D

-glucopyranosyl)-(1fi 2)-a-D-mannopyranosyl-(1fi

3)-[2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi

4)]-[2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi 2)-a-D

-mannopyranosyl-(1fi 6)]-b-D-mannopyranosyl-(1fi

4)-2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi

4)-2-ace-tamido-2-deoxy-b-D-glucopyranoside (4; BiB8AH) A

35.8 mg (16.8 lmol) portion of the protected

octasaccha-ride (3) was dissolved in a mixture of 6.7 mL methanol and

330 lL acetic acid.After addition of 61 mg

palladium-(II)-oxide hydrate (84.6%; Pd), the suspension was stirred under

hydrogen at atmospheric pressure for 24 h (TLC:

propan-2-ol/1Mammonium acetate, 2 : 1, v/v).To drive the reaction

to completion, 50 mg palladium-(II)-oxide hydrate and

300 mL acetic acid were added, and hydrogenation was

continued for 6 h.The catalyst was removed by

centrifu-gation and washed three times with 10% (v/v) acetic acid in

methanol.The combined supernatants were concentrated,

and the remainder (36.0 mg) was purified by gel filtration

(column: Pharmacia Hi Load Superdex 30, 600· 16 mm;

mobile phase: 100 mMNH4HCO3; flow rate: 750 lLÆmin)1; detection: 220 and 260 nm) and lyophilized.The yield was 26.1 mg (95.0%) and analytical data were as follows:

Rf¼ 0.29 (propan-2-ol/1M ammonium acetate, 2 : 1, v/v), ½a22D ¼)1.0  (1.7, H2O) and C64H109N7O41 (M¼ 1632.59) ESI-MS: Mcalcd¼ 1631.68, Mfound¼ 816.9 (M + 2H)2+.(Complete set of 1H/13C-NMR data not shown.)

N1-(6-Aminohexanoylamido)-b-D-galactopyranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi

2)-a-D-mannopyranosyl-(1fi 3)-[2-acetamido-2-deoxy-b-D -glucopyranosyl-(1fi 4)]-[b-D-galactopyranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi 2)-a-D -mannopyranosyl-(1fi 6)]-b-D-mannopyranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi 4)-2-ace-tamido-2-deoxy-b-D-glucopyranoside (5; BiB10AH) A 6.8 mg portion (4.17 lmol) octasaccharide (4) was dissolved

in 2.4 mL 20 mM sodium cacodylate buffer, pH 7.4, containing 1.7 mg BSA, 4.34 lmol NaN3, 2 43 lmol MnCl2, 8 3 mg (12 5 lmol) UDP-galactose, 10.3 U alkaline phosphatase and 206 mU GlcNAc-b1–4-galactosyltrans-ferase (EC 2.4.1.22) The reaction mixture was incubated for

48 h at 37C.After complete reaction (TLC: propan-2-ol/

1M ammonium acetate, 2 : 1, v/v), the precipitate was removed by centrifugation.The supernatant was concen-trated to a volume of 450 lL, purified by gel filtration (column: Pharmacia Hi Load Superdex 30, 600· 16 mm; mobile phase: 100 mMNH4HCO3; flow rate: 750 lLÆmin)1; detection: 220 and 260 nm) and lyophilized.The yield was 7.3 mg (89.4%) and analytical data were as follows:

Rf¼ 0.15 (propan-2-ol/1Mammonium acetate, 2 : 1, v/v),

½a22D ¼ + 2 6  (0.46; H2O) and C76H129N7O51 (M¼ 1956.87) ESI-MS: Mcalcd¼ 1955.79, Mfound¼ 978.9 (M + 2H)2+.(Complete set of 1H/13C-NMR data not shown.)

N1 -(6-Aminohexanoylamido)-(5-acetamido-3,5-dideoxy-a-D-glycero-D-galacto-2-nonulopyranulosonic acid)-(2 fi 6)-b-D-galactopyranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi 2)-a-D -mannopyranosyl-(1fi 3)-[2-acetamido-2-deoxy-b-D -glucopyranosyl-(1fi 4)]-[(5-acetamido-3,5-dideoxy-a-D-glycero-D -gal-acto-2-nonulopyranulosonic acid)-(2fi 6)b-D -galactopy-ranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D -glucopyrano-syl-(1 fi 2)-a-D-mannopyranosyl-(1fi 6)]-b-D -man-nopyranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D -glucopy-ranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D -glucopyrano-side (6; BiB1226AH) A 6.8 mg portion (4.17 lmol) of octasaccharide (4) was dissolved in 2.4 mL 20 mMsodium cacodylate buffer, pH 7.4, containing 1.7 mg BSA, 4.34 lmol NaN3, 2 43 lmol MnCl2, 8.3 mg (12.5 lmol) UDP-galactose, 10.3 U alkaline phosphatase and 206 mU GlcNAc-b1–4-galactosyltransferase.The reaction mixture was incubated for 48 h at 37C.After complete reaction (TLC: propan-2-ol/1M ammonium acetate, 2 : 1, v/v), 7.0 mg (9.1 lmol) CMP-N-acetylneuraminic acid and

75 mU b-galactoside-a2–6-sialyltransferase (EC 2.4.99.1) were added.After the pH had been adjusted to 6.0, incubation at 37C was continued for 24 h.A second portion of 7.0 mg (9.1 lmol) CMP-N-acetylneuraminic acid and 75 mU b-galactoside-a2–6-sialyltransferase was

added (pH 6.0) Incubation at 37C for 24 h was followed

by removal of the precipitate by centrifugation.The

supernatant was concentrated to a volume of 450 lL,

purified by gel filtration (column: Pharmacia Hi Load

Superdex 30, 600· 16 mm; mobile phase: 100 mM

NH4HCO3; flow rate: 750 lLÆmin)1; detection: 220 and

260 nm) and lyophilized The yield was 8.4 mg (79.5%),

and analytical data were as follows: Rf¼ 0.11 (propan-2-ol/

1Mammonium acetate, 2 : 1, v/v),½a22D ¼ + 8 1  (0.28;

H2O) and C98H163N9O67 (M¼ 2539.39) ESI-MS:

Mcalcd¼ 2537.96, Mfound¼ 1270.3 (M + 2H)2+

.(Com-plete set of1H/13C-NMR data not shown.)

N1

-(6-Aminohexanoylamido)-(5-acetamido-3,5-dideoxy-a-D-glycero-D-galacto-2-nonulopyranulosonic

acid)-(2fi 3)-b-D-galactopyranosyl-(1fi

4)-2-acetamido-2-deoxy-b-D-glucopyranosyl-(1fi 2)-a-D

-mannopyranosyl-(1fi 3)-[2-acetamido-2-deoxy-b-D

-glucopyranosyl-(1fi 4)]-[(5-acetamido-3,5-dideoxy-a-D-glycero-D

-gal-acto-2-nonulopyranulosonic acid)-(2fi 3)-b-D

-galacto-pyranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D

-glucopyra-nosyl-(1fi 2)-a-D-mannopyranosyl-(1fi ]-b-D

-manno-pyranosyl-(1fi 4)-2-acetamido-2-deoxy-b-D

-glucopyran-osyl-(1fi 4)-2-acetamido-2-deoxy-b-D-glucopyranoside

(7; BiB1223AH) A 6.8 mg portion (4.17 lmol) of

octasaccharide (4) was dissolved in 2.4 mL 20 mMsodium

cacodylate buffer, pH 7.4, containing 1.7 mg BSA,

4.34 lmol NaN3, 2 43 lmol MnCl2, 8.3 mg (12.5 lmol)

UDP-galactose, 10.3 U alkaline phosphatase and 206 mU

GlcNAc-b1–4-galactosyltransferase.The reaction mixture

was incubated for 48 h at 37C.After complete reaction

(TLC: propan-2-ol/1M ammonium acetate, 2 : 1, v/v),

7.0 mg (9.1 lmol) CMP-N-acetylneuraminic acid and

103 mU b-galactoside-a2–3-sialyltransferase (EC 2.4.99.6)

were added.After the pH had been adjusted to 6.0,

incubation at 37C was continued for 24 h.A second

portion of 7.0 mg (9.1 lmol) CMP-N-acetylneuraminic

acid and 103 mU b-galactoside-a2–3-sialyltransferase were

added (pH 6.0) Incubation at 37C for 24 h was followed

by removal of the precipitate by centrifugation.The

supernatant was concentrated to a volume of 450 lL,

purified by gel filtration (column: Pharmacia Hi Load

Superdex 30, 600· 16 mm; mobile phase: 100 mM

NH4HCO3; flow rate: 750 lLÆmin)1; detection: 220 and

260 nm) and lyophilized The yield was 6.5 mg (61.4%) and

analytical data were as follows: Rf¼ 0.14 (propan-2-ol/1M

ammonium acetate, 2 : 1, v/v), ½a22D ¼ + 11 2  (0.1;

H2O) and C98H163N9O67 (M¼ 2539.39) ESI-MS:

Mcalcd¼ 2537.96, Mfound¼ 1270.3 (M + 2H)2+

.(Com-plete set of1H/13C-NMR data not shown.)

To obtain the neoglycoproteins from the spacered

N-gly-cans, the terminal amino group was converted into a reactive

isothiocyanate [41,42].In detail, a 0.34 lmol portion of each

6-aminohexanoyl-N-glycan (5–7) was dissolved in 200 lL

dilute NaHCO3(100 mg Na2CO3/10 mL H2O) in a 1.5 mL

plastic vessel.A solution of 1 lL (13 1 lmol) thiophosgene in

200 lL dichloromethane was added, and the biphasic

mixture was vigorously stirred.After the amine had been

consumed (1.5 h; TLC: 2-propanol/1Mammonium acetate

2 : 1, v/v; Rfof the decasaccharide derivative¼ 0.45; Rfof

the a2–3-disialylated derivative¼ 0.35; Rf of the

a2–6-disialylated derivative¼ 0.28), the phases were separated by

centrifugation, and the aqueous phase was collected.The organic phase was extracted twice with 100 mL dilute NaHCO3.To remove traces of thiophosgene, the combined aqueous phases were extracted twice with dichloromethane

A 2 mg portion of carbohydrate-free BSA was dissolved in the aqueous solution containing the isothiocyanate deriv-ative.The pH was adjusted to 9.0 by addition of 1MNaOH After 6 days at ambient temperature, the neoglycoprotein was purified by gel filtration (column: Pharmacia Hi Load Superdex 30, 600· 16 mm; mobile phase: 100 mM

NH4HCO3; flow rate: 750 lLÆmin)1; detection: 220 and

260 nm).The product-containing solution was lyophilized

To calculate the extent of oligosaccharide incorporation into the protein carrier, a colorimetric assay was employed [43] Gel electrophoretic analysis under denaturing conditions combined with silver staining of the neoglycoproteins was performed as described [37,38].In addition to these three neoglycoproteins, lactosylated albumin was produced as control by the diazonium and phenylisothiocyanate reac-tions with p-aminophenyl b-lactoside [41,42]

Molecular modeling The simulations were performed on an IBM-SP2 parallel machine using the program DISCOVER on four or eight processors in parallel.Typically, they took several days of CPU time until completed and produced history files with a size ranging from several hundred megabytes to about 2 GB

In detail, molecular dynamics calculations of the N-glycans using parametrization by the CFF91 force field and automatic procedures ofINSIGHT II(Molecular Simulations, San Diego, CA, USA) were run at 1000 K in vacuum using a dielectric constant of four and at 300 K, 400 K and 450 K in

a solvent box with explicit water molecules in the program frame DISCOVER version 2.98 (Accelrys Inc., San Diego, USA).An additional force was applied to the pyranose rings

at 1000 K to avoid ring inversions.After an equilibration period of 100 ps, simulations proceeded for 50–100 ns in the gas phase and for 1–25 ns in the solvent box.To be able to describe the conformational space, which is occupied during the simulation, several characteristic distances between pseudo atoms were evaluated.The pseudo atom co-ordinates

of each sugar moiety are defined by the arithmetic mean value of the co-ordinates of its heavy atoms.Conformational analysis of the data obtained used suitable software tools,

as described previously [44,45]

Lectin-binding assay Purification of galactoside-specific lectins from dried leaves

of mistletoe (Viscum album L) and bovine heart and of the b-galactoside-binding IgG fraction from human serum, the purity controls, biotinylation of the sugar receptors under activity-preserving conditions, and quantitation of carbo-hydrate-specific binding to surface-immobilized neoglyco-proteins have been described in detail previously [37,38,46] The experimental series with increasing concentrations of labeled markers and duplicates at each concentration step were performed at least four times up to saturation of binding, and each data set was algebraically transformed to obtain the Kd value and the number of bound sugar receptor molecules at saturation

Flow cytofluorimetry

Automated flow cytofluorimetric analysis of

carbohydrate-dependent binding of biotinylated marker to the surface of

a panel of human tumor cells of different histogenetic

origin [BLIN-1, pre-B cell line; Croco II, B-lymphoblastoid

cell line; CCRF-CEM, T-lymphoblastoid cell line; K-562,

erythroleukemia cell line; KG-1, acute myelogenous

leukemia cell line; HL-60, promyelocytic cell line;

DU4475, mammary carcinoma cell line; NIH:OVCAR-3,

ovarian carcinoma cell line; C205, SW480, and SW620,

colon adenocarcinoma cell lines; Hs-294T, melanoma cell

line; HS-24, nonsmall cell (epidermoid) lung carcinoma cell

line] using the streptavidin–R-phycoerythrin conjugate as

fluorescent indicator (1 : 40 dilution; Sigma, Munich,

Germany) was performed on a FACScan instrument

(Becton-Dickinson, Heidelberg, Germany) equipped with

the softwareCONSORT30, as described previously [37,38]

To reduce nonspecific binding by protein–protein

interac-tions, cells were incubated with 100 lg ligand-free carrier

protein (BSA) per mL for 30 min at 4C before incubation

with the biotinylated neoglycoprotein in Dulbecco’s

phosphate-buffered saline containing 0.1% (w/v)

ligand-free BSA.The extent of noncarbohydrate-dependent

fluorescence intensity was subtracted in each case from

the total binding

Glycohistochemical processing

Following an optimized procedure for visualizing

carbohy-drate-ligand-dependent binding of the neoglycoproteins to

sections of bronchial tumour [small cell (18 cases) and the

three types of nonsmall cell lung cancer (total number of

cases 60; 20 for each type), mesothelioma (20 cases) and

carcinoid (20 cases)], the specimens were processed under

identical conditions with ABC kit reagents and the

substrates diaminobenzidine/H2O2for development of the

colored, water-insoluble product [42,47,48].A case was

considered to be positive when at least clusters of tumor cells

were specifically stained and the controls excluded binding

of the labeled neoglycoprotein via the protein, the spacer or

the biotin moieties [42,47,48].Also, control experiments

without the incubation step with the marker ruled out

staining by binding of kit reagents, i.e the glycoprotein

avidin or horseradish peroxidase [49]

Biodistribution of radioiodinated neoglycoproteins

Incorporation of125I into the neoglycoproteins to a specific

activity of 11.5 MBqÆ(mg protein))1 was achieved by the

chloramine-T method using limiting amounts of reagents

[50].The retention of radioactivity in organs of

Ehrlich-solid-tumor-bearing ddY mice (7 weeks old; Nihon Clea

Co., Tokyo, Japan) after injection of 28.75 kBq per animal

into the tail vein was determined by a c-counter (Aloka

ARC 300, Tokyo, Japan) and expressed as percentage of the

injected dose per gram of wet tissue or per milliliter of blood

for a group of four mice for each type of neoglycoprotein

and for each time point, as described [37,51,52].The mice

were treated and/or handled according to the Guide

Principles for the Care and Use of Laboratory Animals of

the Japanese Pharmacological Society and with the

approval of Tokyo University of Science’s Institutional Animal Care and Use Committee

Results

Synthetic chemistry

A synthetic biantennary heptasaccharide N-glycan with a single unprotected hydroxy group at the 4¢ position of the b-linked mannose moiety was used to introduce the bisecting GlcNAc residue in a yield of 56%.This reaction pathway was planned as an extension of the basic chemo-enzymatic strategy of N-glycan synthesis and required no change in protecting group manipulations [53,54].After completion of the synthesis of the bisecting octasaccharide, the removal of the base-labile protective groups was straightforward and led to compound 1 (Fig.2).Selective reduction of the azido function at the anomeric center was achieved by the propanedithiol method.After removal of the volatile compounds in high vacuo, the intermediate glycosylamine was coupled to the 6-aminohexanoic acid derivative 2 using standard peptide chemistry activation with TBTU/HOBt which generates the intermediate active ester of 2 (Fig.2).An excess of the activated spacer 2 was required in the coupling step, suggesting that the remaining traces of propane-1,3-dithiol were scavenging the active ester.The purified octasaccharide spacer conjugate 3 was obtained in 54% yield after preparative HPLC.In the final deprotection step, catalytic hydrogenation was used to simultaneously remove the four permanent benzyl groups from the chitobiosyl core and liberate the primary amino function in the spacer part.Compound 4 was easily purified

by gel filtration and processed further to the final carbo-hydrate derivatives 5–7 by enzymatic elongation of the carbohydrate chain using glycosyltransferases.The presence

of alkaline phosphatase ensured removal of inhibitors [55]

In the first enzymatic step, bovine milk b1–4-galactosyl-transferase attached galactosyl moieties to each of the terminal GlcNAc residues of the a1–3 and a1–6 arms.As expected, the bisecting GlcNAc residue was not a substrate, presumably because of sterically blocked access for the enzyme exerted by the two antennae [56].The resulting digalactosylated dodecasaccharide 5 (Fig.2) was purified (89% yield after gel filtration), and portions were subse-quently incubated in a one-pot reaction with CMP-sialic acid and either a2–6 or a2–3-sialyltransferase.After puri-fication, the desired sialylated dodecasaccharides 6 and 7 were obtained in a yield of 80% and 61%, respectively.To ensure the structural identity of all reaction products shown

in Fig.2, compounds 1–7 were routinely analyzed by electrospray ionization MS (detailed in Materials and methods) and by complete assignment of the ring carbons and hydrogens by state-of-the-art 2D NMR techniques (not shown).TOCSY, NOESY, HMQC, HMQC-COSY, HMQC-DEPT and HMQC-TOCSY experiments were employed.Conjugation of the three N-glycans 5–7 to BSA as cytochemically and histochemically inert carrier was acomplished by selective activation of the terminal amino group, as illustrated in Fig.3 The free amines were converted into isothiocyanates by thiophosgene in a bipha-sic reaction, which was followed conveniently by TLC.The isothiocyanates [10 eq.Æ(mol BSA))1] were added directly to

carbohydrate-free BSA and allowed to react for 6 days at

pH 9.0 After purification of the neoglycoproteins,

success-ful conjugation was first visualized by the shift in

electro-phoretic mobility in standard SDS/PAGE (Fig.4) The

mean N-glycan incorporation into the carriers (measured by

a colorimetric assay) for decasaccharide (5) was 3.6

N-glycan molecules per BSA molecule.The reactions with

the sialylated dodecasaccharides 6 and 7 resulted in 4.9 and

3.1 carrier-conjugated glycan chains, respectively Two main

effects of the bisecting GlcNAc unit on the conformation of

the biantennary glycan were pinpointed graphically by

molecular modeling.In comparison with previous work

[6–11], we (a) extended the simulation periods considerably,

(b) added calculations in a solvent box, and (c) assessed

probabilities of population density in the conformational

space with improved statistical reliability and without the

strict dependence on time-averaged distance constraints

Computational chemistry

Computational methods were used to analyze the dynamic

behavior of the a1–3/a1–6 branches in the absence and

presence of the bisecting GlcNAc residue.The calculated

xyz population densities of all monosaccharide building blocks were translated into a strict free-energy grading using the Boltzmann equation.Isocontour plots at a constant free energy level of 1.5 kcalÆmol)1were drawn to visualize the inherent flexibility at each point of the branches and the relation of individual flexibility to structural changes (Fig.5).Interestingly, the molecular dynamics (MD) calcu-lations set to vacuum or a solvent box with water mole-cules gave very similar results.A prevailing influence of van der Waals dispersive and repulsive forces on the conformational properties of this system is consistent with this outcome.Relative to the absence of solvent, the simulated presence of water molecules had a dampening effect on the extent of conformational fluctuations and dynamics of intramolecular mobility.The way in which the bisecting substituent shapes the population density of the biantennary nonasaccharide and decasaccharide is shown in Fig.5B,C.As part of the sugar chain, the bisecting GlcNAc induced separation of the accessed space of the a1–6 arm into two sections (Fig.5C) In comparison with the behavior of the unsubstituted N-glycan, the vicinity of the core trisaccharide becomes fairly accessible for terminal residues of the a1–6 arm, a process referred to as

backfold-Fig 2 Chemical and enzymatic steps to pro-duce galactosylated and sialylated N-glycans substituted with bisecting GlcNAc (a) 1, pro-panedithiol, Et 3 N, MeOH; 2, N-benzyloxy-carbonyl-6-aminohexanoic acid, TBTU, 1-hydroxybenzotriazole (HOBT) (1–2: 54%) (b) Pd-H 2 , AcOH, MeOH (95%).(c) b1–4-galactosyltransferase, UDP-Gal, alkaline phosphatase (89%).(d) a2–6-sialyltransferase, CMP-NeuNAc, alkaline phosphatase (c + d: 80%).(e) a2–3-sialyltransferase, CMP-Neu-NAc, alkaline phosphatase (c + e: 61%).

ing.The wedge-like central GlcNAc moiety accounts for

other changes presented in Fig.5.The inherent flexibility of

the a1–3 arm is clearly restricted in the bisecting compound

(Fig.5B,C).Backfolding and restrained fluctuations have

been confirmed by experimental NMR analysis with model

oligosaccharides [6–11].The contribution of carbohydrate

stacking to chain flexibility becomes apparent when two

energetically favored conformations are scrutinized (Fig.6)

Regions I and II in Fig.6A,B comprise conformations from

the MD trajectories, with distances between pseudo atoms

characteristic of stacking.Examples of the topological

arrangements of the chain constituents from these regions

are illustrated in Fig.6C,D.To compare major aspects of the conformational ensembles of the N-glycans, including the sialylated compounds without/with bisecting GlcNAc, representative snapshots from the MD trajectories are presented in Fig.7 A topological constellation showing how the pronounced flexibility of the a1–6 branch can lead

to a dramatic reduction in intramolecular contact with the bisecting GlcNAc, a situation especially encountered in region III, is depicted in Fig.7C,D

At this point, the frequent occurrence of N-glycosylation should be recalled.It is still difficult to gauge the influence of the protein backbone on glycan flexibility in individual instances.In this context, it is reassuring to note that oligosaccharides from N-glycoproteins, for example the Man5)8 N-glycans from RNase B [57,58], can exhibit conformational behavior similar to that when attached to the protein.Therefore, it can be assumed that constant dynamics or only slightly changed level of mobility will also

be encountered in other cases, especially for neoglycopro-teins with spacer-bound glycans, where spatial constraints exerted by the protein are reduced by adding a linker However, emerging evidence for the varying influence of a bisecting GlcNAc on the processing of branch ends for several glycoproteins precludes general a priori deductions being drawn [59].Nonetheless, Figs 5–7, combined with the experimental evidence reviewed in the Introduction, show how the addition of the bisecting GlcNAc modifies the conformational behavior of the N-glycan.To determine whether binding of sugar receptors to the branches is affected by the bisecting modification, we performed solid-phase assays with the same set of lectins and antibodies, as described previously [37].With galectin-1 as a test substance,

we selected an endogenous lectin that mediates tumor cell adhesion to extracellular matrix glycoproteins, a key step in the metastatic process, invasion of the parenchyma, and growth regulation after ligand cross-linking [27,28,60–67] Solid-phase assay

The incorporation of 3.1–4.9 N-glycan chains per albumin molecule was comparable to the yields in our previous studies [37,38].The neoglycoproteins obtained were thus

Fig 3 Activation of the amino group of the

spacer at the reducing end of the synthetic

N-glycans represented by dodecasaccharide (6)

and the coupling to BSA (a) Thiophosgene,

CH 2 Cl 2 /H 2 O, NaHCO 3 , pH 8.5 (quant.); (b)

BSA, H 2 O, NaHCO 3 , pH 9 0.

Fig 4 Visualization of the gel electrophoretic mobility under denaturing

conditions of carbohydrate-free BSA (A) and the neoglycoproteins

car-rying the decasaccharide with bisecting GlcNAc (5) (B) and the

a2–3-sialylated (C) and a2–6-sialylated (D) derivatives (substances 6 and 7 in

Fig 2) Positions of marker proteins for molecular mass designation

are indicated by arrowheads.

expected to have sufficient ligand density for interaction.In

contrast with natural glycoproteins, the problem of

micro-heterogeneity of the N-glycans is not encountered with these

synthetic products.To monitor the ligand properties

comparatively, we maintained the test profile with a dimeric

plant lectin, mammalian homodimeric galectin-1 and

b-galactoside-specific IgG fraction from human serum

Similar to the interaction of a lectin with a cell surface and

following the protocol of our two previous studies, the

neoglycoproteins were first bound to the surface and the glycan-binding proteins assayed in solution.Thereby, any distortion of the lectin/antibody structures by adsorption to plastic was avoided.In the solid-phase assay, saturable and carbohydrate-dependent binding was measured, and the resulting Scatchard plots gave straight lines indicating the presence of a single class of binding sites in each case (not shown).To allow convenient comparisons, we summarize our data together with previous results on unsubstituted

Fig 5 Illustration of the nomenclature system for the N-glycan constituents (A) and of their inherent flexibility by isocontour plots (derived from analysis of xyz population densities of each monosaccharide unit) at a constant energy level of 1.5 kcalÆmol-1for the biantennary nonasaccharide (B; first structure in Fig 1) and the decasaccharide containing the bisecting GlcNAc (C; structure 4in Fig 1 and substance 5 in Fig 2) For convenient comparison, the conformations were positioned in space in the same way by superimposing the ring atoms of mannose units of the pentasaccharide core.Access to the conformational space in the vicinity of the linear part of the core for the terminal galactose moiety of the a1–6 arm is emphasized

by introducing the term backfolded into the figure (C; also Fig.7C,D).

Fig 6 Analysis of the involvement of the bisecting GlcNAc unit in stacking interactions by measuring pseudo atom distances between this residue (Fig 5A, 9) and spatially neighbouring residues 4, 5 and 4¢, 5¢ of the two branches, respectively, in MD simulations The pseudo atom co-ordinates of each sugar moiety are defined by the arithmetic mean of the co-ordinates of its heavy atoms.Two energetically favored conformational families with glycosidic torsion angle sets of the GlcNAcb1–4Man linkage of F ¼ 50 /Y ¼ 20  (A; global minimum) or F ¼ )30 /Y ¼ )30  (B; side minimum) were separately analyzed.The two illustrated conformations representing examples from regions I and II were taken from the MD trajectories in explicit solvent (for clarity, water molecules are not shown).Distance values between residues 9 and 5 (5¢) located in region I (A) are characteristic of occurrence of stacking indicated by arrows (C).Spatial proximity to the 4 (4¢) residues can still be maintained for the bisecting GlcNAc in region II (A, B, D).In region III, populated by the flexible a1–6 branch, the intramolecular contact with the bisecting GlcNAc is clearly diminished.

N-glycans [37] in Table 1.The general conclusion is that the

presence of a bisecting GlcNAc affects lectin affinity

In the case of mistletoe lectin, the introduction of a

bisecting GlcNAc into the glycan was unfavorable for

binding.The affinity of the galactose-dependent interaction

was about sixfold lower (Bi9-BSA vs.BiB10-BSA, Table 1)

A similar result was obtained for the a2–6-sialylated

derivative, which is a ligand with even greater binding affinity than the sialic-acid-free N-glycan (Bi1126-BSA vs BiB1226-BSA, Table 1) a2–3-Sialylation produced an un-favorable docking site.The introduction of the glycan substituent at a distance from the actual contact site for the lectin can thus indeed modulate ligand properties, without being directly involved in binding.Remarkably, the plant

Fig 7 Illustration of major aspects of the conformational ensembles of sialylated derivatives of the biantennary nonasaccharide Bi9 (A) and of the decasaccharide BiB10 (for nomenclature, see Fig 1) with the bisecting GlcNAc (B–D) The conformations presented were taken from the MD trajectories in explicit solvent (for clarity, water molecules are not shown).The terminal sialic acid moieties of the branches are designated as N and N¢ to allow easy visualization in the extended (A, B) and backfolded structures (C, D).A description of the populated conformational space has been given in Fig.5B,C and that the conformations shown in Fig.6C,D complement the illustration of the extended structure of the biantennary N-glycan with bisecting GlcNAc.Stacking interactions with this moiety (indicated by an arrow between residue 4 in the a1–3 branch as the main contact point and the bisecting GlcNAc) are also possible in backfolded structures (D).

Table 1 Apparent dissociation constants (K d ) for the interaction of (neo)glycoproteins with sugar receptors and the number of bound probe molecules at saturation for Viscum album agglutinin (VAA), bovine galectin-1 and the human b-galactoside-binding IgG subfraction from human serum in a solid-phase assay K d is given in n M ; B max is expressed as bound probe molecules per well.The assays with VAA and the neoglycoprotein BiB1226-BSA were performed with 0.25 lg as matrix.For Lac-BSA (diazo) and Lac-BSA (thio) BSA was glycosylated by covalent attachment of either the diazophenyl derivative (diazo) or the p-isothiocyanatophenyl derivative (thio) of p-aminophenyl b- D -lactoside.Each value is the mean ± SD from

at least four independent experimental series, the quantity of (neo)glycoprotein for coating in lg/well being given for each type of substance.

Matrix

BiB10-BSA (0.5 lg) 163.3 ± 79.4 1.9 ± 0.4 · 10 10

518.6 ± 118 2.1 ± 0.6 · 10 10

73.9 ± 31.6 2.3 ± 1.9 · 10 10 BiB1223-BSA (0.5 lg) 1063 ± 347 1.4 ± 0.6 · 10 10

817.9 ± 493 1.6 ± 0.6 · 10 10

36.1 ± 26.5 2.9 ± 1.1 · 10 10 BiB1226-BSA (0.5 lg) 49.8 ± 19.9 4.2 ± 2.3 · 10 10 829.6 ± 50.6 2.2 ± 0.6 · 10 10 71.8 ± 48.9 2.5 ± 1.1 · 10 10 Bi9-BSA (0.5 lg)a 26.7 ± 11.6 4.6 ± 1.9 · 10 10

900.1 ± 176 42.8 ± 12.5 · 10 10

32.9 ± 19.6 0.35 ± 0.1 · 10 10 Bi1123-BSA (0.5 lg)a 938.4 ± 661 8.2 ± 4.4 · 10 10

829.5 ± 501 42.0 ± 16.5 · 10 10

87.3 ± 62.7 0.38 ± 0.1 · 10 10 Bi1126-BSA (0.5 lg) a 8.7 ± 4.5 6.1 ± 1.4 · 10 10 1025.5 ± 619 48.7 ± 18.4 · 10 10 33.9 ± 4.6 0.46 ± 0.1 · 10 10 Lac-BSA (diazo) (3 lg) 312.4 ± 190 4.7 ± 2.3 · 10 10 1127.2 ± 53.3 34.6 ± 17.6 · 10 10 139.0 ± 87.6 0.70 ± 0.1 · 10 10 Lac-BSA (thio) (0.5 lg) 13.4 ± 7.3 5.1 ± 0.2 · 10 10

516.0 ± 20.3 83.3 ± 6.5 · 10 10

7.6 ± 5.2 0.65 ± 0.1 · 10 10 Asialofetuin (1 lg) 7.4 ± 2.6 4.9 ± 0.5 · 10 10 819.0 ± 268 37.5 ± 10.7 · 10 10 69.2 ± 40.2 0.43 ± 0.1 · 10 10 a

From [37].