Chemoenzymatic Conjugation of Toxic Payloads to the Globally Conserved N-Glycan of Native mAbs Provides Homogeneous and Highly Efficacious Antibody-Drug Conjugates

Article abstract of DOI:10.1021/acs.bioconjchem.5b00224

A robust, generally applicable, nongenetic technology is presented to convert monoclonal antibodies into stable and homogeneous ADCs. Starting from a native (nonengineered) mAb, a chemoenzymatic protocol allows for the highly controlled attachment of any

Full text of DOI:10.1021/acs.bioconjchem.5b00224

Article
pubs.acs.org/bc
Chemoenzymatic Conjugation of Toxic Payloads to the Globally
Conserved N‑Glycan of Native mAbs Provides Homogeneous and
Highly Efficacious Antibody−Drug Conjugates
Remon van Geel, Marloes A. Wijdeven, Ryan Heesbeen, Jorge M. M. Verkade, Anna A. Wasiel,
Sander S. van Berkel, and Floris L. van Delft*
SynAffix BV, Pivot Park, Molenstraat 110, 5342 CC, Oss, The Netherlands
S
* Supporting Information
ABSTRACT: A robust, generally applicable, nongenetic technology is
presented to convert monoclonal antibodies into stable and homogeneous
ADCs. Starting from a native (nonengineered) mAb, a chemoenzymatic
protocol allows for the highly controlled attachment of any given payload to
the N-glycan residing at asparagine-297, based on a two-stage process: first,
enzymatic remodeling (trimming and tagging with azide), followed by
ligation of the payload based on copper-free click chemistry. The technology,
termed GlycoConnect, is applicable to any IgG isotype irrespective of
glycosylation profile. Application to trastuzumab and maytansine, both
components of the marketed ADC Kadcyla, demonstrate a favorable in vitro
and in vivo efficacy for GlycoConnect ADC. Moreover, the superiority of the
native glycan as attachment site was demonstrated by in vivo comparison to a
range of trastuzumab-based glycosylation mutants. A side-by-side comparison
of the copper-free click probes bicyclononyne (BCN) and a dibenzoannu-
lated cyclooctyne (DBCO) showed a surprising difference in conjugation efficiency in favor of BCN, which could be even further
enhanced by introduction of electron-withdrawing fluoride substitutions onto the azide. The resulting mAb-conjugates were in all
cases found to be highly stable, which in combination with the demonstrated efficacy warrants ADCs with a superior therapeutic
index.
payload attachment.10 However, re-engineering of protein
INTRODUCTION
■
sequence and site optimization is a laborious exercise.
More than 50 years after the first mice studies with an
Moreover, expression yields may be compromised (typically
antibody−drug conjugate,1 Ehrlich’s dream2 of the “magic
∼50% yield reduction for a cysteine-engineered mAb,11 while
bullet” for application in the field of oncology was recently
expression yields for genetic encoding of an unnatural amino
realized with two ADCs receiving market approval: Adcetris
acid typically reach 1 g/L maximally),12,13 which has a
(2011), a CD30-targeting ADC for treatment of Hodgkin
significant impact on the cost of goods of the ADC. Finally,
lymphoma and sALCL, and Kadcyla (2013), a HER2-targeting
it has been amply demonstrated that ADCs obtained by
ADC for breast cancer. The promising clinical results of both
conjugation to cysteine side chains often display limited
these marketed ADCs have tremendously boosted interest in
stability in circulation, leading to premature disconnection of
targeted drug delivery for cancer treatment, with currently more
the cytotoxic payload before the tumor site is reached.4 Some
than 40 ADCs in clinical development. The vast majority of
recent improvements in the field are notable,8,9,14 however
ADCs in the clinic are based on conjugation of payload to
obviously still require a cysteine-engineered mAb.
naturally available amino acid side-chains (lysine, cysteine),
Hence, ADC technologies should preferentially meet the
leading to a stochastic distribution of drug−antibody ratio
following criteria: (a) use of native mAb (nongenetic
(DAR) between 0 and 8 (or even higher).3 It has been
approach), (b) control of conjugation site, and (c) delivery of
demonstrated that random conjugation has a negative impact
a highly stable product. We report here a unique and robust
on efficacy,4 and as a consequence the therapeutic index
chemoenzymatic technology for efficient mAb-to-ADC con-
remains low (e.g., efficacious and maximum tolerated dose are
version by anchoring a payload to the antibody’s glycan at
similar for Kadcyla).5,6
Two main strategies can enhance the therapeutic index of a
Special Issue: Antibody-Drug Conjugates
given mAb-payload combination: (a) site-specific conjugation7
and (b) enhancing stability.8,9 Site-specific conjugation is
Received: April 21, 2015
typically achieved by engineering of a specific amino acid (or
Revised: May 25, 2015
sequence) into an antibody, serving as the anchor point for
© XXXX American Chemical Society
A
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
asparagine-297 (Figure 1A). By means of consecutive
enzymatic glycan trimming and tagging of each N-glycan with
antibody monotherapy by binding to Fcγ-receptors,17 thereby
inducing antibody-directed cellular cytotoxicity (ADCC) and/
or complement-dependent cytotoxicity (CDC). For targeting
therapy as with ADCs, however, effector function plays only a
marginal role, whereas the presence of glycans may actually be
considered disadvantageous due to reduced tumor-specific
targeting as a result of uptake by FcγR-positive cells.18 Off-
target cellular uptake of the ADC by mannose receptors may be
additionally disadvantageous.19
We reasoned that the globally conserved glycan may be
rendered advantageous by serving as a natural anchor point for
payloads. However, the heterogeneous nature of the glycan
precludes the direct introduction of a defined number of
payloads onto the glycan. For example, the first reported
approach along this line involved periodate oxidation of
terminal galactose residues, followed by oxime ligation of
payload, which led to a wide distribution of different
conjugates.21 Qasba et al. were the first to demonstrate that,
after prior convergence of different glycoforms to G0(F) upon
galactosidase and sialidase treatment, chemoenzymatic fluo-
rescent labeling of antibody could be effected with galactosyl-
transferase GalT(Y289L) and ketosugar 1 (Figure 2).22,23
Figure 1. (A) Schematic depiction of ADC generation by chemo-
enzymatic remodeling: a native mAb (I, mixture of glycoforms at Asn-
297) is trimmed by an endoglycosidase, thereby liberating the core
GlcNAc (structure II). Next, an azido-modified GalNAc analogue is
attached utilizing a glycosyl transferase (producing III). Third, the
azido-modified glycan is subjected to copper-free click conjugation
with a BCN-modified payload to provide ADC IV. (B) Main
glycosylation forms of recombinant antibodies from CHO (with
indication of abundance).20
Figure 2. Previously reported (1−3) and new (4−5) UDP sugar
substrates for chemoenzymatic remodeling of mAbs.
precisely one azide, a unique anchor point is introduced for
copper-free click conjugation with a payload. The versatility of
this technology, termed GlycoConnect, is demonstrated across
a range of different mAb isotypes and linker−payload
combinations both in vitro and in vivo. Seamless upscaling of
GlycoConnect corroborated manufacturability, whereas the
superiority of the native glycosylation site for conjugation was
demonstrated through in vivo efficacy assessment of a series of
ADCs based on glycosylation mutants.
Lewis et al. also applied a chemoenzymatic strategy to J591, a
PSMA-targeting antibody, with azido-modified GalNAc (Gal-
NAz, 2), followed by radiolabeling with strain-promoted azide−
alkyne cycloaddition (SPAAC).24 Finally, Boons et al. followed
a two-stage approach for ADC generation of a CD22-targeting
antibody via conversion of G0(F) to G2(F) isoforms by
enzymatic galactosylation, followed by enzymatic introduction
of C9-azido-modified sialic acid (SialAz, 3) and subsequent
copper-free click conjugation of payload with a dibenzoannu-
lated cyclooctyne (DIBO).25 The above chemoenzymatic
strategies, by virtue of the branched nature of G0, all generate
ADCs with DAR4, which may not always be desirable, for
example, when extremely potent and/or lipophilic payloads are
employed, e.g., pyrrolobenzodiazepine (PBD) dimers. More-
over, the high lipophilicity of the resulting ADCs, further
enhanced in case a benzoannulated cyclooctyne is applied, may
lead to increased aggregation and accelerated hepatic clearance
from circulation. Finally, it must be noted that mannosylated
glycoforms (typically 5−17%) are not converted into an ADC
based on the above protocols.
RESULTS AND DISCUSSION
■
All monoclonal antibodies are glycosylated at (or around) Asn-
297. While variability of the glycan structure of an antibody
exists within a single batch of mAb (Figure 1B), the exact ratio
of isoforms differs per IgG isotype and is moreover highly
dependent on the mammalian expression system used (CHO,
sp2/0 or NS0).15 Complex glycans (G0−2) typically dominate,
largely fucosylated at the core GlcNAc, while >10% of galactose
residues may additionally be sialylated and various levels of
mannosylated glycans (5−17%) may be present.16 The exact
nature of the glycan is fundamentally important during
B
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
We set out to investigate a chemoenzymatic ligation strategy
to generate DAR2 ADCs, based on the strategy depicted in
Figure 1: (1) trimming of all glycan isoforms (complex, hybrid,
high-mannose) with an endoglycosidase, thereby liberating the
core GlcNAc; (2) enzymatic transfer of a galactose residue
harboring a reactivity-enhanced azide to reduce conjugation
stoichiometry and incubation time; and (3) copper-free click
conjugation with bicyclononyne (BCN), a cyclooctyne with
minimal lipophilicity to reduce aggregation.
Endoglycosidase Treatment. As described above, a mAb
obtained by recombinant expression in a mammalian system
will display high glycan heterogeneity. While the main glycan
isoforms of a mAb expressed in CHO are depicted in Figure
1B, recombinant expression in other mammalian cells (Sp2/0,
NS0) typically further enhances heterogeneity due to formation
of varying amounts of hybrid, bisected, α-galactosylated, and N-
sialylated glycans.16 Thus, effective trimming of all N-glycans
present on any IgG requires access to an endoglycosidase with
excellent substrate tolerability. In addition, high activity of the
endoglycosidase (or more precise: endo-β-N-acetylglucosami-
nidase) is of course also desirable. With this in mind we
selected glycosidases with high glycan tolerability and reported
IgG activity, i.e., endo F3, endo S, and endo S2. As model
antibodies, we used trastuzumab, with a single glycosylation site
at Asn-300, and cetuximab, which belongs to the ∼20% of
mAbs harboring two glycosylation sites (one at Asn-299 and
one at Asn-88).26
antibody (i.e., structure III in Figure 1 labeled with galactose
derivative 2 as depicted in Figure 2, denoted as III-2).
Click Conjugation/Choice of Cyclooctyne. Having
successfully obtained bis-GalNAz-modified trastuzumab (III-
2), the stage was set to investigate the conjugation of a
functional group by means of click chemistry. Our initial
investigations focused on the use of copper-mediated click
chemistry. To this end, trastuzumab (GalNAz)2 III-2 was
incubated with commercially available propargyl (+)-biotin
derivative 6 (Figure 3), in the presence of CuSO4 and Na-
Endo F3 is known to cleave a wide variety of complex
glycans, but not oligomannose and hybrid structures and has
400-fold lower activity on nonfucosylated glycans.27 Indeed, we
found that upon incubation of trastuzumab with endo F3, the
majority of glycans was readily removed, but full trimming was
not attained, even with large excess of enzyme. A similar
situation occurred with endo S,28,29 which is highly active
even with nonfucosylated glycansbut fails to hydrolyze
oligomannose structures. Finally, we anticipated that endo S2
would display the desirable combination of substrate tolerability
and activity.30 Indeed, incubation of trastuzumab with only 1%
(w/w) of enzyme was sufficient for full hydrolysis of the
glycans, including oligomannose structures, as judged by SDS-
PAGE and mass spectrometric analysis (Supporting Informa-
tion).
Galactosyl Transferase. In 2002, Qasba et al. reported that
a single mutation of a wild-type galactosyltransferase, GalT-
(Y289L), resulted in excellent transferase activity of N-acetyl-D-
galactosamine (GalNAc)31 and could also be applied to transfer
a 2-keto-derivative of galactose (1, Figure 2) onto a GlcNAc
substrate.32 For the highly similar polypeptide-galactosaminyl-
transferase, Bertozzi et al. showed that the azido-modified
GalNAc analogue GalNAz (2, Figure 2) is an effective
substrate.33 These observations were elegantly combined by
Lewis et al. in 2013, showing that GalT(Y289L) is also able to
recognize and incorporate GalNAz onto glycoproteins.24 We
speculated that GalT(Y289L) may not only be effective in
functionalization of the natural substrate for galactosylation
the terminal GlcNAc of complex glycansbut also be applied
in transfer of GalNAz onto the mAb core N-GlcNAc liberated
upon endoglycosidase treatment (structure II in Figure 1).
Indeed, we found that incubation of trimmed trastuzumab (II)
with GalNAz 2 (0.4 mM) in the presence of only 1% (w/w) of
GalT(Y289L) and MnCl2 (10 mM) led to complete labeling
after overnight incubation, thereby effectively installing
precisely two azide groups (one per heavy chain) onto the
Figure 3. (A) Structures of propargyl-(+)-biotin (6), BCN-PEGn-
pyrene (7a−c), and DBCO-PEGn-pyrene (8a−c) constructs. (B)
Conjugation efficiency of cyclooctynes, depending on PEG chain
length, for 7 and 8.
ascorbate (Supporting Information). While complete labeling
was observed after overnight incubation, as corroborated by MS
analysis, a close inspection of the mass profile of the adduct, in
particular after enzymatic digestion, indicated that several
amino acids (His-288, Met-4) had suffered from partial
concomitant oxidation. Although careful optimization of
reaction conditions could possibly annihilate undesired amino
acid oxidation, it was reasoned that such optimization would
need to take place on case-by-case (mAb) basis, since even
small amounts of oxidized protein may induce immunogenic
response. An elegant solution to avoid copper involves train-
promoted chemistry with cyclooctynes, as originally applied by
Bertozzi et al.34 However, the low reaction rate constants of the
original cyclooctynes urged the development of a whole range
of improved cyclooctynes over the years.35 Some of us (S.V.B.,
F.V.D.) contributed to the field (2010) bicyclo[6.1.0]nonyne
(BCN)36 and dibenzoazacyclooctyne (DBCO, originally named
DiBAC),37 two cyclooctynes with a significantly improved
reactivity profile, which have nowadays firmly established
themselves in the field of strain-promoted azide−alkyne
cycloaddition.
We anticipated that copper-free conjugation would fully
avoid the aforementioned undesired amino acid oxidation.
Thus, we prepared a range of derivatives (Figure 3A) of BCN-
C
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
Table 1. Overview of Conditions and Conjugation Efficiencies of Azido-Modified Trastuzumab Variants III-2, III-4, and III-5
a
with a Range of BCN-Linker Payload Constructs
entry
UDP-sugar
linker
payload
equiv
cosolvent
DMF
amount (%)
conversion (%)
1
GalNAz (2)
PEG8
doxorubicin
MMAE
2
5
5
>95
>95
>95
90
2
PEG2-glut-vc-PABA
PEG2-glut-vc-PABA
PEG4-Ahx
5
3
MMAF
1.5
2
5
4
maytansine
maytansine
duocarmycin SA
MMAF
5
5
PEG2-vc-PABA-β-Ala
PEG4-vc-PABA-DMEDA
PEG2-glut-vc-PABA
PEG4-Ahx
2.5
1.5
4
25
12.5
5
60
6
85
7
F2−GalNAz (4)
F2−GalNBAz (5)
DMSO
DMA
DMF
DMA
DMA
>95
>95
>95
>95
>95
>95
>95
8
maytansine
maytansine
duocarmycin SA
maytansine
maytansine
MMAF
3
5
9
PEG2-vc-PABA-β-Ala
PEG4-vc-PABA-DMEDA
PEG4-Ahx
3
5
10
11
12
13
3
2.5
5
3
PEG2-vc-PABA-Ahx
PEG2-glut-vc-PABA
3
5
4
DMSO
5
a
Full structures of the Linker-Payloads can be found in the Supporting Information. PEG = polyethylene glycol, glut = glutaryl, vc = Val-Cit, PABA =
p-aminobenzylalcohol, Ahx = 6-aminohexanoyl, DMEDA = N,N′-dimethylethylene diamine, MMAE = monomethyl auristatin E, MMAF =
monomethyl auristatin F.
pyrene (7) and DBCO-pyrene (8), separated by a range of
ethylene glycol linkers of different length (n = 4, 8, 12). Pyrene
was selected to mimic the typically hydrophobic payloads
needed for ADC purpose, while the PEG-linkers of different
chain length were included to evaluate their potential to
facilitate conjugation of the hydrophobic pyrene and cyclo-
octyne moieties. In addition, we compared conjugation
efficiency of both BCN and DBCO under aqueous conditions,
in particular, in light of the anticipated higher reactivity of
DBCO versus BCN (for aliphatic azides), but also due to the
much greater hydrophobic nature of the former. Hence,
trastuzumab(GalNAz)2 III-2 (Figure 3) was incubated with
7a−c or 8a−c and progress of the reactions was monitored by
mass spectrometry (Supporting Information). As anticipated,
conjugation of BCN-PEG-pyrene gradually improved by
increasing PEG chain length from 4 to 8 ethylene glycol
units (no further improvement was observed for PEG12). A
similar trend was noted for conjugation of DBCO-PEGn-pyrene
(n = 4 or 8); however, much to our surprise further increase in
ethylene glycol chain length to 12 in fact thwarted the
conjugation process, which did not exceed beyond 20%.38 The
inability of DBCO to reach full conversion during SPAAC has
in fact already been reported for labeling of monoclonal
antibodies containing an azidoamino acid.13 Nonetheless, in all
cases it was found that SPAAC with BCN (reaching full
conversion) was clearly favorable over DBCO (leveling off at
max ∼60%). In addition, we established that ADCs containing
DBCO show enhanced propensity to aggregate compared to
their BCN-analogues (Supporting Information).
highly hydrophobic BCN-PEG4-vc-PABA-DMEDA-duocarmy-
cin SA construct (entry 6). Shortening of the PEG linker, as in
the BCN-PEG2-glutaryl derivative (entry 5), not only required
25% of cosolvent, but also resulted in reduced conjugation
efficiency (60%). It must be noted that, although all
conjugations in Table 1 were performed on trastuzumab as
model system, application to other mAbs afforded similar
results (vide infra). Stability studies were performed by
incubating ADCs based on MMAF and maytansine for a
prolonged time in blood plasma (2 weeks, 37 °C), which
showed no sign of degradation of linker in either case.40 In
addition, aggregation studies under similar conditions showed
<1% formation of aggregates after 2 weeks in plasma at 37 °C
for both ADCs (see Supporting Information). Our next
objective was to demonstrate the scalability of the Glyco-
Connect technology. To this end, 5 g of Rituxan (rituximab),
an antibody known to be less stable than trastuzumab, was
subjected to the chemoenzymatic modification protocol. Much
to our satisfaction, we found that upscaling GlycoConnect to
increasing amounts of rituximab (5 mg → 100 mg → 1 g → 5
g) and subsequent conjugation with BCN-PEG8-doxorubicin
proceeded seamlessly, affording the desired ADC with excellent
homogeneity (Supporting Information).
In Vitro and In Vivo Evaluation. Some of the ADCs
prepared above were evaluated for their cell-killing potential on
a Her2-overexpressing cell line (SK-Br-3) and compared to a
Her2-negative cell line (MDA-MB231). As becomes clear from
Figure 4, both maytansine- and MMAF-based ADCs showed
ADC Production and Stability. With all components in
place (enzymes, substrate, and BCN), the stage was set to
generate ADCs via our glycan remodeling−copper-free click
strategy. Hence, we synthesized a range of BCN-linker-payload
constructs, summarized in Table 1 (structures summarized in
Supporting Information), with variation of linker (short/long,
cleavable/noncleavable) and payload (monomethyl auristatins
E and F, duocarmycin SA, maytansine).39 As becomes apparent
from Table 1 (entries 1−6) varying amounts of organic
cosolvent were required, depending on linker-payload, to
ensure full solubility during conjugation. In the majority of
cases, 5% of DMF sufficed to ensure full conjugation (>95%) of
BCN-linker-payload, although 12.5% was required for the
Figure 4. In vitro cytotoxicity of GlycoConnect ADCs based on vc-
PABA-maytansine (blue line) and vc-PABA-MMAF (green line)
versus Kadcyla (red line) and trastuzumab (black line).
D
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
comparable or better cytotoxicity than the benchmark Kadcyla,
the marketed ADC that is also based on trastuzumab and
maytansine hence only differs in conjugation technology.
The advantage of glycan-remodeled ADCs versus randomly
labeled ADCs becomes even more apparent in vivo. To this
end, PDX mice with an average tumor size of 100 mm3 were
intravenously injected with one of three different GlycoCon-
nect ADCs (cleavable/noncleavable with maytansine or
cleavable with MMAF) and compared to Kadcyla at 9 mg/kg
(single dose bolus injection), followed by twice weekly tumor
sizes measurements (Figure 5). During the first week, all groups
azides.42 For example, it was demonstrated that an azidopyr-
idinium derivative reacts >30 times faster than benzyl azide
with BCN (but not with DBCO) following an inverse electron-
demand mechanism. We speculated that a similar effect may be
operative upon introduction of electron-withdrawing fluoride
substituents on aliphatic azides, to potentially further enhance
payload conjugation efficiency. Indeed, a competition experi-
ment between model substrates N3CH2C(O)NHBn and
N3CF2C(O)NHBn for reaction with BCN indicated a 2.3-
fold higher activity of the difluorinated analogue (see
Supporting Information). Based on this result, we designed
and synthesized the difluorinated UDP-GalNAc derivative
UDP-F2−GalNAz (4, Figure 2) for evaluation in GlycoCon-
nect. Enzymatic incorporation with GalT(Y289L) onto
trimmed trastuzumab II (Figure 1) proceeded smoothly
under standard conditions, to produce trastuzumab (F2−
GalNAz)2 derivative III-4. Another competition experiment
was performed, in this case involving incubation of equimolar
amounts of nonfluorinated (GalNAz-based) III-2 and difluori-
nated III-4 with a substoichiometric amount of BCN-PEG2000
in PBS buffer, again indicating a >2-fold increase in reactivity
for the difluorinated analogue (see Supporting Information).
Encouraged by these results, we investigated whether GalT-
(Y289L) would also be able to incorporate the azide derivative
of difluorinated N-benzoyl-D-galactosamine (F2−GalNBAz, 5,
Figure 2), which based on our earlier study was expected to
display a reaction rate constant at least 15 times higher than
GalNAz. Initial remodeling experiments, under standard
conditions, indicated only poor enzymatic incorporation of
F2−GalNBAz. However, increasing the concentration of UDP-
substrate 5 to 5 mM and GalT(Y289L) to 15% (w/w) ensured
full conversion of II, leading to quantitative formation of
trastuzumab(F2−GalNBAz) III-5 as judged by mass spectrom-
etry. In order to improve the activity of the galactosyltransfer-
ase, we designed and produced a small set of Y289 mutants.
Interestingly, we found that one of these mutants, GalT-
(Y289M), dramatically enhanced the efficiency of incorporation
of F2−GalNBAz, which ensured full conversion into III-5 under
much improved conditions (150 μM GalT(Y289M) and 5 mM
5).
With two difluorinated azido-trastumumab derivatives in
hand, the stage was set to evaluate their suitability of ADC
preparation. Thus, III-4 (F2−GalNAz) or III-5 (F2−GalNBAz)
were incubated with BCN-Ahx-maytansine or BCN-vc-PABA-
maytansine), with minimal reagent stoichiometry and low
reaction times. In line with the enhanced activities of the
respective azides, it was found that full conversion of III-4/5
into IV-4/5 could be ensured within 4 h using as little as 1.5
equiv of BCN-linker-toxin per azide. The enhanced reactivities
of the difluorinated substrates may translate into significantly
improved manufacturability (reduced stoichiometry and time in
plant) and, hence, cost of goods.
Having generated most of our data with trastuzumab, a
particular robust monoclonal antibody, we next moved to
different monoclonal antibodies to demonstrate the applic-
ability of GlycoConnect. First of all, we investigated the
applicability of GlycoConnect on IgG2 and IgG4 isotypes. Both
antibody formats are known to be susceptible to domain
scrambling under reducing conditions, which are inevitable for
any cysteine-based conjugation technology. Second, we
evaluated GlycoConnect in the context of mAbs produced in
Sp2/0 and NS0. A range of commercially available therapeutic
antibodies was obtained: infliximab (IgG1 from Sp2/0),
Figure 5. In vivo efficacy of GlycoConnect in a mouse HBCx-15 PDX
xenograft with single injection (9 mg/kg) of trastuzumab-based ADCs.
Full and permanent regression becomes apparent with maytansine-
based ADCs, both with cleavable (Val-Cit-p-aminobenzylalcohol [vc-
PABA], red line) or noncleavable linker (6-aminohexyl [Ahx], yellow
line). Some tumor regrowth becomes apparent with vc-PABA-MMAF
(green line), which is more prominent with Kadcyla (blue line). Black
line is vehicle.
showed significant tumor shrinkage or even complete
regression. After 1 week, tumor sizes gradually increased for
mice treated with Kadcyla but not for any of the GlycoConnect
ADCs. Only after approximately 3 weeks, reappearance and
regrowth of tumors became apparent for mice treated with
trastuzumab-vc-PABA-MMAF, while both ADCs based on
maytansine (cleavable or noncleavable) showed complete
tumor regression for the whole length of the study (60 days).
This study thus clearly demonstrates the superiority of
GlycoConnect conjugation technology versus a randomly
conjugated ADC based on the same components. The latter
observation becomes particularly interesting in view of the
higher drug loading of Kadcyla (average DAR3.5)3 versus
GlycoConnect ADCs (DAR2).
Finally, it was established that glycan remodeling does not
negatively impact pharmacokinetics. Although theory dictates
that the mAb glycan plays no role in reshuttling of mAb by the
FcRn receptor,41 we felt it important to corroborate the latter
also in the context of glycan-remodeled ADCs. To this end,
trastuzumab (native and trimmed) and both trastuzumab-vc-
PABA-based ADCs were (randomly) radiolabeled with 111In
and injected into mice. Samples were taken at preset time
intervals and measured for residual radioactivity. As expected,
no difference in blood clearance was observed (Supporting
Information).
More Reactive Azido-Substrates. Recently, it was
reported by van Delft et al. that reaction rate constants for
SPAAC with BCN can be tremendously accelerated by applying
electron-poor aromatic azides instead of the traditional aliphatic
E
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
palivizumab (IgG1 from NS0), panitumumab (IgG2 from
CHO), and natalizumab (IgG4 from NS0). After dialysis to
remove excipients, each mAb was subjected to our standard
protocol for conversion into ADC and the process was followed
by mass spectrometry. In all cases, clean and quantitative
conversion into ADC was observed (Figure 6).
Figure 7. In vivo site-scanning of five different glycomutants
demonstrating superior efficacy for the native glycosylation site (red
line) compared to all glycosylation mutants included.
Figure 6. Mass spectral data (crude) of glycan-remodeling−
conjugation of different IgG isotypes. Top row displays mass spectral
profile of the heavy chain of native mAb, middle line after glycan
remodeling (with F2−GalNAz 4), bottom spectrum after conjugation
with a BCN-maytansine construct.
linker−payload combination. The resulting ADCs were found
to be homogeneous and highly hydrolytically stable, while
displaying negligible aggregation. Benchmark in vitro and in
vivo efficacy studies against the marketed product Kadcyla,
based on the same antibody and payload components,
generated strong biological data in favor of the GlycoConnect
ADCs, despite the lower DAR. The superiority of the natural
mAb glycosylation site around Asn-297 was demonstrated in
vivo by a side-by-side comparison of a range of trastuzumab
glycomutants. Current efforts in our laboratory are focusing on
the development of linkers with increased potential to
accommodate highly lipophilic payloads (duocarmycins, PBD
dimers) versus the commonly employed ethylene glycol linkers,
which show only limited usefulness in this regard. Moreover,
we are looking into branched linker formats that may allow for
ADCs with higher DAR (4−8), and possibly even combina-
tions of payloads with different modes of action, to combat
multidrug-resistant tumors with a combination therapy-like
approach. In summary, GlycoConnect technology is readily
applicable to any off-the-shelf monoclonal antibody (turn-
around in <1 week), while delivering site-specific, stable, and
highly efficacious ADCs. We are confident therefore that
GlycoConnect shows high promise as the next-generation ADC
generation technology for targeted therapy with improved
therapeutic index.
Site-Scanning Study. Finally, we set out to investigate if
the native glycosylation site is indeed optimal in the context of
efficacy. It has been amply demonstrated by others that there is
a relevant correlation between conjugation site and ADC
efficacy.43 Most recently, it was demonstrated44 that ADCs with
cleavable linkers are significantly less stable in mouse plasma if
not conjugated at or around Asn-297. The latter observation is
highly inconvenient if a mouse xenograft model is used to
assess ADC efficacy, although (presumably) not prohibitive for
eventual clinical application in humans. Endorsing the
importance of the site of conjugation, we performed a small
site-scanning study for trastuzumab-based ADCs by engineer-
ing a new glycosylation site into trastuzumab heavy chain or
light chain (based on the sequence N-X-S/T) and simulta-
neously mutating Asn-300 to Gln (trastuzumab numbering).
The newly designed trastuzumab double mutants were
transiently expressed in CHO and glycosylation was confirmed
by mass spectrometry. Next, each individual trastuzumab
mutant was subjected to endoglycosidase endo F3,45 followed
by enzymatic incorporation of F2−GalNAz 4, purification by
protein A affinity chromatography, and SPAAC with BCN-
PEG4-Ahx-maytansine. The resulting, noncleavable ADCs were
purified by size-exclusion chromatography and the resulting
ADCs were injected into PDX mice at suboptimal dose (6 mg/
kg) in order to be able to quantify differences in efficacy. We
were pleased to find that, in accordance with previous reports,
the most efficacious ADC (red line) was indeed derived from
native trastuzumab Asn-300 (Figure 7). However, an important
difference is that the set of ADCs investigated in our study
involved solely noncleavable linkers, which excludes the role of
a potential endogenous serine hydrolase in differentiating
between the range of conjugates.44
EXPERIMENTAL PROCEDURES
■
Solvents were purchased from Sigma-Aldrich or Fisher
Scientific and used as received. Thin layer chromatography
was performed on silica gel-coated plates (Kieselgel 60 F254,
Merck, Germany) with the indicated solvent mixture, spots
were detected by KMnO4 staining (1.5 g KMnO4, 10 g K2CO3,
2.5 mL 5% NaOH-solution, 150 mL H2O), p-anisaldehyde
staining (9.2 mL p-anisaldehyde, 321 mL EtOH, 17 mL H2O,
3.75 mL AcOH, 12.7 mL H2SO4), and UV-detection. NMR
spectra were recorded on a Bruker Biospin 400 (400 MHz) and
a Bruker DMX300 (300 MHz). Protein mass spectra (HRMS)
were recorded on a JEOL AccuTOF JMS-T100CS (Electro-
spray Ionization (ESI) time-of-flight) or a JEOL AccuTOF
JMS-100GCv (Electron Ionization (EI), Chemical Ionization
(CI)). Low-resolution mass spectra (LRMS) were recorded on
a ThermoScientific Advantage LCQ Linear ion-trap electro-
spray or a Waters LCMS consisting of a 2767 Sample manager,
2525 pump, 2996 UV-detector, and a Micromass ZQ with an
Xbridge C18 3.5 μm column (ESI).
CONCLUSION
■
A straightforward and robust technology for conversion of
mAbs into ADCs has been developed, based on enzymatic
remodeling of naturally present glycans, followed by copper-
free click conjugation with BCN-payload constructs. The
methodology is generally applicable to any recombinant IgG
isotype, irrespective of mammalian expression system and
F
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
3,4,6-Tri-O-acetyl-2-azido-2-deoxy-α-D-galactose-1-
phosphate (9). Compound 9 was prepared from D-galactos-
amine according to the procedure described for D-glucosamine
by Linhardt et al.46 1H NMR (300 MHz, CD3OD): δ 5.69 (dd,
J = 7.2, 3.3 Hz, 1H), 5.43−5.42 (m, 1H), 5.35 (dd, J = 11.1, 3.3
Hz, 1H), 4.53 (t, J = 7.2 Hz, 1H), 4.21−4.13 (m, 1H), 4.07−
4.00 (m, 1H), 3.82 (dt, J = 10.8, 2.7 Hz, 1H), 2.12 (s, 3H), 2.00
(s, 3H), 1.99 (s, 3H). LRMS (ESI-) calcd for C12H17N3O11P
(M-H−) 410.06, found 410.00.
(s, 3H). LRMS (ESI−) calcd for C23H29F2N6O20P2 (M-H−)
809.09, found 809.1.
UDP 2-(2′-Azido-2′,2′-difluoroacetyl)-α-D-galactos-
amine (UDP-F2−GalNAz, 4). Compound 12 was dissolved
in H2O (1 mL) and triethylamine (1 mL) and MeOH (2.4 mL)
were added. The reaction mixture was stirred for 2 h and then
concentrated in vacuo. Flash chromatography (7:2:1−5:2:1
EtOAc:MeOH:H2O) afforded UDP F2−GalNAz (4; Scheme
1
1). H NMR (300 MHz, D2O): δ 7.86 (d, J = 8.1 Hz, 1H),
3,4,6-Tri-O-acetyl-α-D-galactosamine-1-phosphate
(10). To a solution of azide 9 (105 mg, 0.26 mmol) in MeOH
(3 mL) was added Pd/C (20 mg). The reaction was stirred
under a hydrogen atmosphere for 2 h and filtered over Celite.
The filter was rinsed with MeOH (10 mL) and the filtrate was
concentrated in vacuo to afford free amine 10 (94 mg, 0.244
Scheme 1. Synthesis of UDP-Sugars 4 and 5
1
mmol, 96%). H NMR (300 MHz, D2O): δ 5.87−5.76 (m,
1H), 5.44 (br s, 1H), 5.30−5.20 (m, 1H), 4.55 (t, J = 6.3 Hz,
1H), 4.28−4.00 (m, 3H), 2.11 (s, 3H), 2.03 (s, 3H), 2.00 (s,
3H). LRMS (ESI−) calcd for C12H19NO11P (M-H−) 384.07,
found 384.10.
3,4,6-Tri-O-acetyl-N-(2′-azido-2′,2′-difluoroacetyl)-α-
D-galactosamine-1-phosphate (11). To a solution of amine
10 (94 mg, 0.244 mmol) in dry DMF (3 mL) were added ethyl
difluoroazidoacetate (48 mg, 0.293 mmol) and Et3N (68 μL,
0.49 mmol). The reaction was stirred for 6 h at r.t., followed by
concentration in vacuo to afford the crude product. Flash
chromatography (100:0−50:50 EtOAc:MeOH) afforded prod-
5.91−5.85 (m, 2H), 5.54 (dd, J = 6.6, 3.6 Hz, 1H), 4.31−3.95
(m, 9H), 3.74−3.62 (m, 2H). LRMS (ESI−) calcd for
C17H23F2N6O17P2 (M-H−) 683.06, found 683.10.
1
uct 11 (63 mg, 0.125 mmol, 51%). H NMR (300 MHz,
CD3OD): δ 5.64 (m, 1H), 5.47 (d, J = 2.4 Hz, 1H), 5.35 (dd, J
= 11.4, 3.0 Hz, 1H), 4.58−4.48 (m, 2H), 4.25−4.15 (m, 1H),
4.09−4.00 (m, 1H), 2.14 (s, 3H), 2.00 (s, 3H), 1.93 (s, 3H).
LRMS (ESI-) calcd for C14H18F2N4O12P (M-H−) 503.06, found
503.0.
UDP 3,4,6-Tri-O-acetyl-2-(2′-azido-2′,2′-difluoroace-
tyl)-)-α-D-galactosamine (12). Monophosphate 11 was
coupled to UMP according to the procedure reported by
Baisch et al.47 Thus, a solution of D-uridine-5′-monophosphate
disodium salt (98 mg, 0.266 mmol) in H2O (1 mL) was treated
with DOWEX 50Wx8 (H+ form) for 40 min and filtered. The
filtrate was stirred vigorously at r.t. while tributylamine (63 μL,
0.266 mmol) was added dropwise. After 30 min of stirring, the
reaction mixture was lyophilized and further dried over P2O5
under vacuum for 5 h.
UDP 3,4,6-Tri-O-acetyl-2-azido-2-deoxy-α-D-galactose
(13). To a solution of D-uridine-5′-monophosphate disodium
salt (1.49 g, 4.05 mmol) in H2O (15 mL) was treated with
DOWEX 50Wx8 (H+ form) for 30 min and filtered. The filtrate
was stirred vigorously at room temperature while tributylamine
(0.966 mL, 4.05 mmol) was added dropwise. After 30 min of
further stirring, the reaction mixture was lyophilized and further
dried over P2O5 under vacuum for 5 h.
The resulting tributylammonium uridine-5′-monophosphate
was dissolved in dry DMF (25 mL) under an argon
atmosphere. Carbonyldiimidazole (1.38 g, 8.51 mmol) was
added and the reaction mixture was stirred at r.t. for 30 min.
Next, dry MeOH (180 μL) was added and the reaction was
stirred for 15 min to remove the excess carbonyldiimidazole.
The leftover MeOH was removed under high vacuum for 15
min. Subsequently, compound 9 (2.0 g, 4.86 mmol) was
dissolved in dry DMF (25 mL) and added dropwise to the
reaction mixture. The reaction was allowed to stir at r.t. for 2 d
before concentration in vacuo. The consumption of the
imidazole-UMP intermediate was monitored by MS. Flash
chromatography (7:2:1−5:2:1 EtOAc:MeOH:H2O) afforded
The resulting tributylammonium uridine-5′-monophosphate
was dissolved in dry DMF (15 mL) under an argon
atmosphere. 1,1′- Carbonyldiimidazole (35 mg, 0.219 mmol)
was added and the reaction mixture was stirred at r.t. for 30
min. Next, dry MeOH (4.63 μL) was added and the reaction
was stirred for 15 min to remove the excess carbonyl
diimidazole. The leftover MeOH was removed under high
vacuum for 15 min. Subsequently, N-methylimidazole, HCl salt
(61 mg, 0.52 mmol) was added to the reaction mixture and the
monophosphate 11 (63 mg, 0.125 mmol) was dissolved in dry
DMF (15 mL) and added dropwise to the reaction mixture.
The reaction was allowed to stir at r.t. overnight before
concentration in vacuo. The consumption of the imidazole-
UMP intermediate was monitored by MS. Flash chromatog-
raphy (7:2:1−5:2:1 EtOAc:MeOH:H2O) afforded the product
1
product 13 (1.08 g, 1.51 mmol, 37%). H NMR (300 MHz,
D2O): δ 7.96 (d, J = 8.0 Hz, 1H), 5.98−5.94 (m, 2H), 5.81−
5.79 (m, 1H), 5.70 (dd, J = 7.1, 3.3 Hz, 1H), 5.49 (dd, J = 15.2,
2.6 Hz, 1H), 5.30 (ddd, J = 18.5, 11.0, 3.2 Hz, 2H), 4.57 (q, J =
6.0 Hz, 2H), 4.35−4.16 (m, 9H), 4.07−3.95 (m, 2H), 2.17 (s,
3H), 2.08 (s, 3H), 2.07 (s, 3H). LRMS (ESI−) calcd for
C21H29N5O19P2 (M-H−) 716.09, found 716.3.
UDP 2-Azido-2-deoxy-α-D-galactose (14). Compound
13 (222 mg, 0.309 mmol) was dissolved in H2O (2.5 mL) and
triethylamine (2.5 mL) and MeOH (6 mL) were added. The
reaction mixture was stirred for 3 h at r.t. and then
1
12. H NMR (300 MHz, D2O): δ 7.87 (d, J = 8.1 Hz, 1H),
5.913−5.85 (m, 2H), 5.67 (dd, J = 6.6, 2.7 Hz, 1H), 5.56−5.50
(m, 1H), 5.47−5.43 (m, 1H), 5.31−5.25 (m, 2H), 4.61−4.43
(m, 2H), 4.31−4.05 (m, 5H), 2.16 (s, 3H), 2.02 (s, 3H), 1.94
1
concentrated in vacuo to afford crude 14. H NMR (300
MHz, D2O): δ 7.99 (d, J = 8.2 Hz, 1H), 6.02−5.98 (m, 2H),
G
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
5.73 (dd, J = 7.4, 3.4 Hz, 1H), 4.42−4.37 (m, 2H), 4.30−4.18
(m, 4H), 4.14−4.04 (m, 2H), 3.80−3.70 (m, 2H), 3.65−3.58
(m, 1H). LRMS (ESI−) calcd for C15H23N5O16P2 (M-H−)
590.05, found 590.2.
After deconvolution of peaks, the mass spectrum showed one
peak of the light chain and two peaks of the heavy chain. The
two peaks of heavy chain belong to one major product (49 496
Da, 90% of total heavy chain), resulting from core GlcNAc-
(Fuc) substituted trastuzumab, and a minor product (49 351
Da, 10% of total heavy chain), resulting from trimmed
trastuzumab lacking L-fucose.
Protocol for Glycosyltransfer of UDP-GalNAz (2) or
UDP-F2−GalNAz (4) with GalT(Y289L) to Obtain III-2 or
III-4. Trimmed trastuzumab II (10 mg/mL) was incubated with
the appropriate UDP-GalNAc derivative 2 (commercially
available from CarboSynth, Compton, Berkshire, UK) or 4
(0.4 mM) and β(1,4)-Gal-T1(Y289L)48 (0.1 mg/mL) in 10
mM MnCl2 and 25 mM Tris-HCl pH 8.0 for 16 h at 30 °C.
Next, the functionalized trastuzumab was purified on a HiTrap
MabSelect SuRe 5 mL column (GE Healthcare) using an
AKTA purifier-10 (GE Healthcare). The eluted IgG was
immediately neutralized with 1.5 M Tris-HCl pH 8.8 and
dialyzed against PBS pH 7.4. Next the IgG was concentrated
using an Amicon Ultra-0.5, Ultracel-10 Membrane (Millipore)
to a concentration of 20 mg/mL.
Mass spectral analysis of III-2 after reduction with DTT
indicated the formation of one major product (49 713 Da, 90%
of total heavy chain), resulting from GalNAz transfer to core
GlcNAc(Fuc) substituted trastuzumab, and a minor product
(49 566 Da, 10% of total heavy chain), resulting from GalNAz
transfer to core GlcNAc substituted trastuzumab.
Mass spectral analysis of III-4 after reduction with DTT
indicated the formation of one major heavy chain product
(49 865 Da, approximately 90% of total heavy chain), resulting
from F2−GalNAz transfer to core GlcNAc(Fuc)-substituted
trastuzumab which has subsequently reacted with DTT during
sample preparation.
Preparation of Trastuzumab-(F2−GalNBAz), III-5.
Trimmed trastuzumab II (10 mg/mL) was incubated with
UDP F2−GalNBAz 5 (5 mM) and β(1,4)-Gal-T1(Y289M) (1
mg/mL) in 10 mM MnCl2 and 25 mM Tris-HCl pH 8.0 at 30
°C overnight. Mass spectral analysis of a DTT-reduced sample
indicated the formation of a one major product (49 813 Da,
approximately 90% of total heavy chain), resulting from F2−
GalNBAz transfer to core GlcNAc(Fuc) substituted trastuzu-
mab heavy chain.
General Protocol for Conjugation of Cyclooctyne
Constructs (BCN/DBCO) to azido-mAbs III. Three volumes
of a solution of the appropriate trastuzumab(azide)2 III (20
mg/mL in PBS pH 7.4) were added to one volume of a
solution containing the BCN/DBCO compound (≥1.2 mM
solution in Milli-Q and/or organic cosolvent). Analytical scale
reactions were performed at 10 μL scale and samples were
taken at preset time points. Mass spectral analysis was
performed as described above to determine the conjugation
efficiency (see Table 1).
UDP α-D-Galactosamine (15). To a solution of compound
14 in H2O:MeOH 1:1 (4 mL) was added Lindlar’s catalyst (50
mg). The reaction was stirred under a hydrogen atmosphere for
5 h and filtered over Celite. The filter was rinsed with H2O (10
mL) and the filtrate was concentrated in vacuo to afford 15
1
(169 mg, 0.286 mmol, 92% yield over two steps). H NMR
(300 MHz, D2O): δ 7.93 (d, J = 8.1 Hz, 1H), 5.99−5.90 (m,
2H), 5.76−5.69 (m, 1H), 4.39−4.34 (m, 2H), 4.31−4.17 (m,
5H), 4.05−4.01 (m, 1H), 3.94−3.86 (m, 1H), 3.82−3.70 (m,
3H), 3.30−3.16 (m, 1H). LRMS (ESI−) calcd for
C15H25N3O16P2 (M-H−) 564.06, found 564.1.
UDP 2-(4′-Azido-3′,5′-difluorobenzoyl)-α-D-galactos-
amine (UDP F2−GalNBAz, 5). To a solution of 4-azido-3,5-
difluorobenzoic acid was added dicyclohexylcarbodiimide (1.1
equiv) and N-hydroxysuccinimide (1.2 equiv) and the resulting
suspension was stirred overnight at r.t. followed by vacuum
filtration. The filtrate was concentrated and dissolved in EtOAc
followed by washing with saturated NaHCO3 (aq.) and brine.
The organic layer was dried over Na2SO4, filtrated, and
1
concentrated in vacuo to use crude in the next reaction. H
NMR (300 MHz, CDCl3): δ 7.74−7.66 (m, 2H), 2.91 (s, 4H).
Next, UDP GalNH2 (15, 30 mg, 0.0531 mmol) was dissolved
in 0.1 M NaHCO3 (0.2 M) and the N-hydroxysuccinimide
ester of 4-azido-3,5-difluorobenzoic acid (31 mg, 0.106 mmol, 2
equiv), dissolved in DMF (0.2 M), was added. The reaction
was stirred overnight at r.t. and concentrated in vacuo. Flash
chromatography (7:2:1−5:2:1 EtOAc:MeOH:H2O) afforded
1
product 5 (8 mg, 0.011 mmol, 20%). H NMR (300 MHz,
D2O): δ 7.73 (d, J = 8.4 Hz, 1H), 7.52−7.31 (m, 2H), 5.87−
5.71 (m, 2H), 5.65−5.57 (m, 1H), 5.47−5.33 (m, 1H), 4.43−
3.96 (m, 8H), 3.76−3.60 (m, 2H). LRMS (ESI−) calcd for
C22H25F2N6O17P2 (M-H−) 745.07, found 744.9.
General Protocol for Mass Spectral Analysis of IgG.
Prior to mass spectral analysis, IgGs were either treated with
DTT, which allows analysis of both light and heavy chain, or
treated with Fabricator (commercially available from Genovis,
Lund, Sweden), which allows analysis of the Fc/2 fragment. For
analysis of both light and heavy chain, a solution of 20 μg
(modified) IgG was incubated for 5 min at 37 °C with 100 mM
DTT in a total volume of 4 μL. If present, azide-functionalities
are reduced to amines under these conditions. For analysis of
the Fc/2 fragment, a solution of 20 μg (modified) IgG was
incubated for 1 h at 37 °C with Fabricator (1.25 U/μL) in
phosphate-buffered saline (PBS) pH 6.6 in a total volume of 10
μL. After reduction with DTT or Fabricator-digestion the
samples were washed trice with Milli-Q using an Amicon Ultra-
0.5, Ultracel-10 Membrane (Millipore) resulting in a final
sample volume of approximately 40 μL. Next, the samples were
analyzed by electrospray ionization time-of-flight (ESI-TOF)
on a JEOL AccuTOF. Deconvoluted spectra were obtained
using Magtran software.
Preparative scale reactions were purified on a Superdex 200
10/300 GL (GE Healthcare) using an AKTA purifier-10 (GE
Healthcare).
Glycan Trimming of Trastuzumab with Endo S.
Trastuzumab I (10−20 mg/mL) was incubated with endo S
from Streptococcus pyogenes (4 U/mg IgG, commercially
available as IgGZERO from Genovis, Lund, Sweden) in 25
mM Tris pH 8.0 for 16 h at 37 °C. Next, trimmed IgG (II) was
concentrated and washed with 10 mM MnCl2 and 25 mM Tris-
HCl pH 8.0 using an Amicon Ultra-0.5, Ultracel-10 Membrane
(Millipore).
ASSOCIATED CONTENT
* Supporting Information
Experimental procedures for the syntheses of compounds 7a−c
and 8a−c, conjugation of 6−8 to trastuzumab(GalNAz)2 and
analysis, preparation and analysis of ADCs, protocols for in
vitro and in vivo studies, and competition experiments. The
Supporting Information is available free of charge on the ACS
■
S
H
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
(10) Aggerwal, P., and Bertozzi, C. R. (2014) Site-specific antibody−
drug conjugates: the nexus of bioorthogonal chemistry, protein
engineering, and drug development. Bioconjugate Chem. 53, 176−92.
(11) For example, an extensive optimization process to avoid
formation of triple light chain antibody format of a THIOmab did not
afford titers above 2 g/L, whereas for regular IgG yields of 3−5 g/L are
customary. Gomez, N., Ouyang, J., Nguyen, M. D. H., Vinson, A. R.,
Lin, A. A., and Yuk, I. H. (2010) Effect of temperature, pH, dissolved
oxygen, and hydrolysate on the formation of triple light chain
antibodies in cell culture. Biotechnol. Prog. 26, 1438−45.
(12) Tian, F., Lu, Y., Manibusan, A., Sellers, A., Tran, H., Sun, Y.,
Phuong, T., Barnett, R., Hehli, B., Song, F., DeGuzman, M. J., Ensari,
S., Pinkstaff, J. K., Sullivan, L. M., Biroc, S. L., Cho, H., Schultz, P. G.,
DiJoseph, J., Dougher, M., Ma, D., Dushin, R., Leal, M., Tchistiakova,
L., Feyfant, E., Gerber, H. P., and Sapra, P. (2014) A general approach
to site-specific antibody drug conjugates. Proc. Natl. Acad. Sci. U. S. A.
111, 1766−71.
Publications website at DOI: 10.1021/acs.bioconj-
chem.5b00224.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: F.vanDelft@SynAffix.com.
Notes
The authors declare the following competing financial
interest(s): all authors are or have been (except A.A.W.)
employees of SynAffix B.V.
ACKNOWLEDGMENTS
■
The work presented here is partially supported by a Stichting
Technische Wetenschappen (STW) Valorization Grant
(12305). Louis Boon from Bioceros (Utrecht, The Nether-
lands) is acknowledged for providing trastuzumab-containing
supernatant.
(13) Zimmerman, E. S., Heibeck, T. H., Gill, A., Li, X., Murray, C. J.,
Madlansacay, M. R., Tran, C., Uter, N. T., Yin, G., Rivers, P. J., Yam, A.
Y., Wang, W. D., Steiner, A. R., Bajad, S. U., Penta, K., Yang, W.,
Hallam, T. J., Thanos, C. D., and Sato, A. K. (2014) Production of site-
specific antibody−drug conjugates using optimized non-natural amino
acids in a cell-free expression system. Bioconjugate Chem. 25, 351−61.
(14) Fontaine, S. D., Reid, R., Robinson, L., Ashley, G. W., and Santi,
D. V. (2014) Long-term stabilization of maleimide−thiol conjugates.
Bioconjugate Chem. 26, 145−52.
REFERENCES
■
(1) Mathe,
́
G., Loc, T. B., and Bernard, J. (1958) Effet sur la leucem
hopterine
ie par
́
ie
1210 de la souris d’une combinaison par diazotation d’A-met
́
́
et de γ-globulines de hamsters porteurs de cette leucem
heter
́
́ ́
ogreffe. Compt. Rendu Acad. Sci. (Paris) 246, 1626−8.
(2) Himmelweit, F. (1960) The Collected Papers of Paul Ehrlich, p 59,
(15) Beck, A., Wagner-Rousset, E., Ayoub, D., Van Dorsselaer, A.,
Vol. 3, Pergamon, London.
́
and Sanglier-Cianferani, S. (2013) Characterization of therapeutic
(3) Chari, R. V. J., Miller, M. L., and Widdison, W. C. (2014)
Antibody−Drug Conjugates: An Emerging Concept in Cancer
Therapy. Angew. Chem., Int. Ed. 53, 3796−827.
antibodies and related products. Anal. Chem. 85, 715−36.
(16) Goetze, A. M., Liu, Y. D., Zhang, Z., Shah, B., Lee, E.,
Bondarenko, P. V., and Flynn, G. C. (2011) High-mannose glycans on
the Fc region of therapeutic IgG antibodies increase serum clearance in
humans. Glycobiology 21, 949−59.
(4) Junutula, J. R., Raab, H., Clark, S., Bhakta, S., Leipold, D. D.,
Weir, S., Chen, Y., Simpson, M., Tsai, S. P., Dennis, M. S., Lu, Y.,
Meng, Y. G., Ng, C., Yang, J., Lee, C. C., Duenas, E., Gorrell, J., Katta,
V., Kim, A., McDorman, K., Flagella, K., Venook, R., Ross, S., Spencer,
S. D., Wong, W. L., Lowman, H. B., Vandlen, R., Sliwkowski, M. X.,
Scheller, R. H., Polakis, P., and Mallet, M. (2008) Site-specific
conjugation of a cytotoxic drug to an antibody improves the
therapeutic index. Nat. Biotechnol. 26, 925−32.
(17) Ferrara, C., Grau, S., Jager, C., Sondermann, P., Brunker, P.,
̈
̈
Waldhauer, I., Hennig, M., Ruf, A., Rufer, A. C., Stihle, M., Umana, P.,
̃
and Benz, J. (2011) Unique carbohydrate-carbohydrate interactions
are required for high affinity binding between FcgammaRIII and
antibodies lacking core fucose. Proc. Natl. Acad. Sci. U. S. A. 108,
12669−74.
(5) Krop, I. E., Beeram, M., Modi, S., Jones, S. F., Holden, S. N., Yu,
W., Girish, S., Tibbitts, J., Yi, J.-H., Sliwkowski, M. X., Jacobson, F.,
Lutzker, S. G., and Burris, H. A. (2010) Phase I study of trastuzumab-
DM1, an HER2 antibody-drug conjugate, given every 3 weeks to
patients with HER2-positive metastatic breast cancer. J. Clin. Oncol. 28,
2698−704.
(6) Beeram, M., Krop, I. E., Burris, H. A., Girish, S. R., Yu, W., Lu, M.
W., Holden, S. N., and Modi, S. (2012) A phase 1 study of weekly
dosing of trastuzumab emtansine (T-DM1) in patients with advanced
human epidermal growth factor 2−positive breast cancer. Cancer 118,
5733−40.
(18) For example, see discussion in Sazinsky, S. L., Ott, R. G., Silver,
N. W., Tidor, B., Ravetch, J. V., Wittrup, K. D. (2008) Aglycosylated
immunoglobulin G1 variants productively engage activating Fc
receptors. Proc. Natl. Acad. Sci. U.S.A. 105, 20167−72.
(19) Gorovits, B., and Krinos-Fiorotti, C. (2013) Proposed
mechanism of off-target toxicity for antibody-drug conjugates driven
by mannose receptor uptake. Cancer Immunol. Immunother. 62, 217−
23.
(20) Up to 5−7% of nonglycosylated mAb may also be present,
depending on protein sequence and expression system.
(21) Qu, Z., Sharkey, R. M., Hansen, H. J., Shih, L. B., Govindan, S.
V., Shen, J., Goldenberg, D. M., and Leung, S. (1998) Carbohydrates
engineered at antibody constant domains can be used for site-specific
conjugation of drugs and chelates. J. Immunol. Meth. 213, 131−44.
(22) Boeggeman, E., Ramakrishnan, B., Pasek, M., Manzoni, M., Puri,
A., Loomis, K. H., Waybright, T. J., and Qasba, P. K. (2009) Site
specific conjugation of fluoroprobes to the remodeled Fc N-glycans of
monoclonal antibodies using mutant glycosyltransferases: application
for cell-surface antigen detection. Bioconjugate Chem. 20, 1228−36.
(23) Zhu, Z., Ramakrishnan, B., Li, J., Wang, Y., Feng, Y., Prabakaran,
P., Colantonio, S., Dyba, M. A., Qasba, P. K., and Dimitrov, D. S.
(2014) Site-specific antibody-drug conjugation through an engineered
glycotransferase and a chemically reactive sugar. mAbs 6, 1−11.
(24) Zeglis, B. M., Davis, C. B., Aggeler, R., Kang, H.-C., Chen, A.,
Agnew, B., and Lewis, J. S. (2013) An enzyme-mediated methodology
for the site-specific radiolabeling of antibodies based on catalyst-free
click chemistry. Bioconjugate Chem. 24, 1057−67.
(7) Shen, B.-Q., Xu, K., Liu, L., Raab, H., Bhakta, S., Kenrick, M.,
Parsons-Reponte, K. L., Tien, J., Yu, S.-F., Mai, E., Li, D., Tibbitts, J.,
Baudys, J., Saad, O. M., Scales, S. J., McDonald, P. J., Hass, P. E.,
Eigenbrot, C., Nguyen, T., Solis, W. A., Fuji, R. N., Flagella, K. M.,
Patel, D., Spencer, S. D., Khawli, L. A., Ebens, A., Wong, W. L.,
Vandlen, R., Kaur, S., Sliwkowski, M. X., Scheller, R. H., Polakis, P.,
and Junutula, J. R. (2012) Conjugation site modulates the in vivo
stability and therapeutic activity of antibody-drug conjugates. Nat.
Biotechnol. 30, 184−9.
(8) Lyon, R. P., Setter, J. R., Bovee, T. D., Doronina, S. O., Hunter, J.
H., Anderson, M. E., Balasubramanian, C. L., Duniho, S. M., Leiske, C.
I., Li, F., and Senter, P. D. (2014) Self-hydrolyzing maleimides
improve the stability and pharmacological properties of antibody-drug
conjugates. Nat. Biotechnol. 32, 1059−1062.
(9) Tumey, L. N., Charati, M., He, T., Sousa, E., Ma, D., Han, X.,
Clark, T., Casavant, J., Loganzo, F., Barletta, F., Lucas, J., and Graziani,
E. I. (2014) Mild method for succinimide hydrolysis on ADCs: impact
on ADC potency, stability, exposure, and efficacy. Bioconjugate Chem.
25, 1871−80.
(25) Li, X., Fang, T., and Boons, G.-J. (2014) Preparation of well-
defined antibody−drug conjugates through glycan remodeling and
I
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX

Bioconjugate Chemistry
Article
strain-promoted azide−alkyne cycloadditions. Angew. Chem., Int. Ed.
53, 7179−82.
S., Wong, T., Chin, S. M., Yu, J., Dilley, J., Chaparro-Riggers, J., Filzen,
G. F., O’Donnell, C. J., Wang, F., Myers, J. S., Pons, J., Shelton, D. L.,
and Rajpal, A. (2013) Location matters: site of conjugation modulates
stability and pharmacokinetics of antibody drug conjugates. Chem. Biol.
20, 161−7.
(44) Dorywalska, M., Strop, P., Melton-Witt, J. A., Hasa-Moreno, A.,
Farias, S. E., Galindo Casas, M., Delaria, K., Lui, V., Poulsen, K., Loo,
C., Krimm, S., Bolton, G., Moine, L., Dushin, R., Tran, T.-T., Liu, S.-
H., Rickert, M., Foletti, D., Shelton, D. L., Pons, J., and Rajpal, A.
(2015) Effect of attachment site on stability of cleavable antibody drug
conjugates. Bioconjugate Chem., DOI: 10.1021/bc5005747.
(45) Neither endo S nor endo S2 could be used in this case since
both are only hydrolytically active on the native glycosylation site of
IgGs.
(26) Exact numbering of amino acids differs per mAb.
(27) For an overview of different endo F glycosidase, see for
example: http://www.sigmaaldrich.com/technical-documents/
articles/biology/glycobiology/endoglycosidases.html (accessed on
April 20th, 2015). Also, see the following review: Karamanos, Y.
(2013) Endo-N-acetyl-β-D-glucosaminidases and peptide-N4-(N-
acetyl-β-D-glucosaminyl) asparagineamidases: More than just tools.
Adv. Biochem. 1, 81−99.
(28) Colin, M., and Olsen, A. (2001) EndoS, a novel secreted protein
from Streptococcus pyogenes with endoglycosidase activity on human
IgG. EMBO J. 20, 3046−55.
(29) Goodfellow, J. J., Baruah, K., Yamamoto, K., Bonomelli, C.,
Krishna, B., Harvey, D. J., Crispin, M., Scanlan, C. N., and Davis, B. G.
(2012) An endoglycosidase with alternative glycan specificity allows
broadened glycoprotein remodelling. J. Am. Chem. Soc. 134, 8030−3.
(46) Masuko, S., Bero, S., Green, D. E., Weïwer, M., Liu, J.,
DeAngelis, P. L., and Linhardt, R. J. (2012) Chemoenzymatic
synthesis of uridine diphosphate-GlcNAc and uridine diphosphate-
GalNAc analogs for the preparation of unnatural glycosaminoglycans.
J. Org. Chem. 77, 1449−56.
(30) Sjogren, J., Struwe, W. B., Cosgrave, E. F. J., Rudd, P. M.,
̈
Stervander, M., Allhorn, M., Hollands, A., Nizet, V., and Collin, M.
(2013) EndoS2 is a unique and conserved enzyme of serotype M49
group A Streptococcus that hydrolyses N-linked glycans on IgG and α1-
acid glycoprotein. Biochem. J. 455, 107−18.
̈
(47) Baisch, G., and Ohrlein, R. (1997) Convenient chemoenzymatic
synthesis of β-purine-diphosphate sugars (GDP-fucose-analogues).
Bioorg. Med. Chem. 5, 383−91.
(48) Boeggeman, E. E., Ramakrishnan, B., and Qasba, P. K. (2003)
The N-terminal stem region of bovine and human β-1,4-galactosyl-
transferase I increases the in vitro folding efficiency of their catalytic
domain from inclusion bodies. Protein Expression Purif. 30, 219−29.
(31) Ramakrishnan, B., and Qasba, P. K. (2002) Structure-based
design of β1,4-galactosyltransferase I (β4Gal-T1) with equally efficient
N-acetylgalactosaminyltransferase activity. J. Biol. Chem. 277, 20833−9.
(32) Khidekel, N., Arndt, S., Lamarre-Vincent, N., Lippert, A., Poulin-
Kerstien, K. G., Ramakrishnan, B., Qasba, P. K., and Hsieh-Wilson, L.
C. (2003) A chemoenzymatic approach toward the rapid and sensitive
detection of O-GlcNAc posttranslational modifications. J. Am. Chem.
Soc. 125, 16162−3.
(33) Hang, H. C., Yu, C., Pratt, M. R., and Bertozzi, C. R. (2004)
Probing glycosyltransferase activities with the Staudinger ligation. J.
Am. Chem. Soc. 126, 6−7.
(34) Agard, N. J., Prescher, J. A., and Bertozzi, C. R. (2004) A strain-
promoted [3 + 2] azide-alkyne cycloaddition for covalent modification
of biomolecules in living systems. J. Am. Chem. Soc. 126, 15046−7.
(35) Debets, M. F., van Berkel, S. S., Dommerholt, J., Dirks, A. J.,
Rutjes, F. P. J. T., and van Delft, F. L. (2011) Bioconjugation with
strained alkenes and alkynes. Acc. Chem. Res. 44, 805−15.
(36) Dommerholt, J., Schmidt, S., Temming, R., Hendriks, L. J.,
Rutjes, F. P. J. T., van Hest, J. C. M., Lefeber, D. J., Friedl, P., and van
Delft, F. L. (2010) Readily accessible bicyclononynes for bioorthog-
onal labeling and three-dimensional imaging of living cells. Angew.
Chem., Int. Ed. 49, 9422−5.
(37) Debets, M. F., van Berkel, S. S., Schoffelen, S., Rutjes, F. P. J. T.,
van Hest, J. C. M., and van Delft, F. L. (2010) Aza-
dibenzocyclooctynes for fast and efficient enzyme PEGylation via
copper-free (3 + 2) cycloaddition. Chem. Commun. 46, 97−9.
(38) We have no explanation for this unexpected observation at this
point, but we speculate that back-folding of dibenzocyclooctyne onto
the polyaromatic pyrene may potentially take place, thereby precluding
the desired copper-free click reaction.
(39) Detailed synthesis and characterization can be found in the
Supporting Information.
(40) Only degradation of maytansine, predominantly ester
hydrolysis, was observed.
(41) Jung, S. T., Reddy, S. T., Kang, T. H., Borrok, M. J., Sandlie, I.,
Tucker, P. W., and Georgiou, G. (2010) A glycosylated IgG variants
expressed in bacteria that selectively bind FcγRI potentiate tumour cell
killing by monocyte-dendritic cells. Proc. Natl. Acad. Sci. U.S.A. 201,
604−9.
(42) Dommerholt, J., van Rooijen, O., Borrmann, A., Fonseca
Guerra, C., Bickelhaupt, F. M., and van Delft, F. L. (2014) Highly
accelerated inverse electron-demand cycloaddition of electron-
deficient azides with aliphatic cyclooctynes. Nat. Commun. 5, DOI:
10.1038/ncomms6378.
(43) Strop, P., Liu, S. H., Dorywalska, M., Delaria, K., Dushin, R. G.,
Tran, T. T., Ho, W. H., Farias, S., Casas, M. G., Abdiche, Y., Zhou, D.,
Chandrasekaran, R., Samain, C., Loo, C., Rossi, A., Rickert, M., Krimm,
J
DOI: 10.1021/acs.bioconjchem.5b00224
Bioconjugate Chem. XXXX, XXX, XXX−XXX