Perfluoroalkyl chains direct novel self-assembly of insulin

Article abstract of DOI:10.1021/la203042c

The self-assembly of biopharmaceutical peptides into multimeric, nanoscale objects, as well as their disassembly to monomers, is central for their mode of action. Here, we describe a bioorthogonal strategy, using a non-native recognition principle, for control of protein self-assembly based on intermolecular fluorous interactions and demonstrate it for the small protein insulin. Perfluorinated alkyl chains of varying length were attached to desB30 human insulin by acylation of the ε-amine of the side-chain of LysB29. The insulin analogues were formulated with Zn^II and phenol to form hexamers. The self-segregation of fluorous groups directed the insulin hexamers to self-assemble. The structures of the systems were investigated by circular dichroism (CD) spectroscopy and synchrotron small-angle X-ray scattering. Also, the binding affinity to the insulin receptor was measured. Interestingly, varying the length of the perfluoroalkyl chain provided three different scenarios for self-assembly; the short chains hardly affected the native hexameric structure, the medium-length chains induced fractal-like structures with the insulin hexamer as the fundamental building block, while the longest chains lead to the formation of structures with local cylindrical geometry. This hierarchical self-assembly system, which combines Zn^II mediated hexamer formation with fluorous interactions, is a promising tool to control the formation of high molecular weight complexes of insulin and potentially other proteins.

Full text of DOI:10.1021/la203042c

ARTICLE
pubs.acs.org/Langmuir
Perfluoroalkyl Chains Direct Novel Self-Assembly of Insulin
Leila Malik,† Jesper Nygaard,† Rasmus Hoiberg-Nielsen,† Lise Arleth,†,* Thomas Hoeg-Jensen,‡,*
and Knud J. Jensen*,†
†IGM, Faculty of Life Sciences, University of Copenhagen, Thorvaldsensvej 40, 1871 Frederiksberg, Denmark
‡Diabetes Protein and Peptide Chemistry, Novo Nordisk, Park D6.1.142, 2760 Maaloev, Denmark
S Supporting Information
b
ABSTRACT: The self-assembly of biopharmaceutical peptides into multimeric, nanoscale objects,
as well as their disassembly to monomers, is central for their mode of action. Here, we describe a
bioorthogonal strategy, using a non-native recognition principle, for control of protein self-
assembly based on intermolecular fluorous interactions and demonstrate it for the small protein
insulin. Perfluorinated alkyl chains of varying length were attached to desB30 human insulin by
acylation of the ε-amine of the side-chain of LysB29. The insulin analogues were formulated with
ZnII and phenol to form hexamers. The self-segregation of fluorous groups directed the insulin
hexamers to self-assemble. The structures of the systems were investigated by circular dichroism
(CD) spectroscopy and synchrotron small-angle X-ray scattering. Also, the binding affinity to the
insulin receptor was measured. Interestingly, varying the length of the perfluoroalkyl chain
provided three different scenarios for self-assembly; the short chains hardly affected the native
hexameric structure, the medium-length chains induced fractal-like structures with the insulin
hexamer as the fundamental building block, while the longest chains lead to the formation of
structures with local cylindrical geometry. This hierarchical self-assembly system, which combines ZnII mediated hexamer formation
with fluorous interactions, is a promising tool to control the formation of high molecular weight complexes of insulin and potentially
other proteins.
’ INTRODUCTION
insulin to form hexamers and the ability to use organic chemistry to
modify insulin has led to different approaches to obtain long-acting
insulin. One approach has been acylation of the ε-amino group of
residue LysB29 with a long-chain saturated fatty acid to promote
binding to circulating serum albumin after injection.10,11 Another
strategy was the use of N-lithocholyl insulin analogues.12 These
analogues proved to be multihexamers that gave a prolonged effect.
The higher-ordered self-assembled structures were dominant in
formulations without phenol (insulin T-form).
Here, we present a bioorthogonal approach to create molec-
ular recognition between proteins. The addition of perfluori-
nated chains to organic molecules changes their physical and
chemical properties,13 as highly fluorinated molecules tend to
segregate into a fluorous phase, a phenomenon which is often
referred to as a fluorous effect.13 The fluorous effect has been
used to direct folding intramolecularly in peptides and proteins,
by incorporation of amino acids with fluorinated side-chains.14 In
contrast, placing a perfluoroalkyl group on the surface of a
protein could potentially direct intermolecular fluorous self-
assembly, which would be a new and bio-orthogonal approach
to control the oligomeric state of proteins and enable the
formation of soluble macromolecular complexes. Applied to
The oligomeric state of peptide and protein based pharma-
ceuticals is essential for their efficacy.1 One way to address this is
to design protein systems that self-organize spontaneously in a
well-defined manner by self-assembly under controlled condi-
tions. Insulin is a clear illustration of this, since monomeric
insulin is fast-acting, whereas insulin variants that predominantly
form insulin hexamers and higher-ordered oligomers are longer-
acting. Insulin is most commonly given as subcutaneous injec-
tions that form a depot and the biophysical properties of these
depots can be exploited to tune insulin delivery.2,3 One way to
construct higher-order oligomer insulin assemblies using non-
covalent insulin hexamers would be to attach ligands that self-
assemble in a bioorthogonal manner to form supramolecular
species. This could lead to new prolonged-action insulins that
provide improved blood glucose control.
Human insulin consists of two peptide segments, the A- and
B-chains with 21 and 30 amino acids, respectively, which are con-
nected by two disulfide bridges; a third disulfide bridge is found in
the A-chain.4 DesB30 insulin (1), which is human insulin without
the C-terminal Thr (at position 30 in the B-chain), is the insulin
backbone used in insulin detemir.5 Insulin exists as a hexamer in the
pancreas as well as in drug formulations. The hexameric form
consists of a trimer of dimers via coordination of two ZnII atoms.
Insulin exists in two hexameric forms known as the T- and R-
conformations. The R-conformer is promoted in solution by the
presence of phenol, m-cresol or resorcinol.6ꢀ9 This propensity of
Received: August 4, 2011
Revised:
November 15, 2011
Published: November 30, 2011
r
2011 American Chemical Society
593
dx.doi.org/10.1021/la203042c Langmuir 2012, 28, 593–603
|

Langmuir
ARTICLE
insulin, this method could lead to a slow release of insulin ana-
logues into the bloodstream and thereby prolong their effect.
In the present project, we chemically modified insulin with
perfluoroalkyl chains of chain-length C3F7 to C9F19. The insulin
analogous were then formulated as for ordinary human insulin
and other analogues. The secondary structures of the formed
aggregates were investigated by means of circular dichroism
(CD) spectroscopy while the 3D low resolution structure, hence
the topology, were investigated by synchrotron small-angle X-ray
scattering.
26.2 (CH2), 25.3 (CH2ꢀCH2), 25.2 (CH2). MS, calculated average
isotopic composition for C14H8F15NO4, 539.0 Da; found: m/z 561.9
[M+Na].
2,5-Dioxopyrrolidine-1-yl 4,4,5,5,6,6,7,7,8,8,9,9,9-tridecafluorono-
nanoate 4. Compound 4 was prepared by reaction of 2H,2H,3H,3H-
perfluorononanoic acid (500 mg, 1.28 mmol) with TSTU (461 mg,
1.2 mmol) and DIEA (261 μL, 1.53 mmol) overnight in THF (3 mL)
1
according to the general procedure, yield 238 mg (38%). H NMR
CDCl3 δ 2.9 (2H, t), 2.8 (4H, s), 2.59ꢀ2.41 (2H, m). 13C NMR
C(D3)2SO δ 168.7 (CdO amide), 166.7 (CdO ester), 26.2 (CH2),
25.6 (CH2ꢀCH2), 22.8 (CH2). MS, calculated average isotopic com-
position for C13H8F15NO4, 489.0 Da; found: m/z 511.9 [M+Na].
2,5-Dioxopyrrolidine-1-yl 4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,11-hep-
tadecafluoroundecanoate 5. Compound 5 was prepared by reaction
of 2H,2H,3H,3H-perfluoroundecanoic acid (492.1 mg, 1 mmol) with
TSTU (361 mg, 1.2 mmol) and DIEA (205 μL, 1.2 mmol) overnight in
THF (3 mL) according to the general procedure, yield 241 mg (41%).
’ EXPERIMENTAL METHODS
All organic solvents were obtained from Iris Biotech, Germany.
DesB30 human insulin was from Novo Nordisk, Denmark. All fluori-
nated reagents were obtained from Fluorous Technologies Inc., USA.
Analytical HPLC was performed on a Dionex Ultimate 3000 with
Chromeleon 6.80SP3 software. Insulin analogues were analyzed on a
Phenomenex Gemini 110 Å C18 column (3 μm, 4.6 ꢁ 50 mm) and on a
Phenomenex Gemini 110 Å C4 column (3 μm, 4.6 ꢁ 50 mm), applying
a flow of 1.0 mL/min with a linear gradient with increasing amount of
buffer B over 10ꢀ20 min. Analytical method 1: buffer A: 0.1% formic
acid in H2O; buffer B: 0.1% formic acid in CH3CN. Analytical method 2:
buffer C: 10 mM phosphate in H2O; buffer D: 10% 100 mM phosphate
in CH3CN, pH 7.4. Preparative HPLC was performed on a Dionex
Ultimate 3000 with Chromeleon 6.80SP3 software. Insulin analogues
were purified on a FEF 300 Å C4 column (5 μm, 20 ꢁ 250 mm),
applying a flow of 10.0 mL/min with a linear gradient with increasing
amount of buffer (10% B to 100% B) over 37 min (buffer A: 0.1% TFA in
H2O; buffer B: 0.1% TFA in CH3CN). Mass analysis was performed on
an ESI-MS Mass Spectrometer (MSQ Plus, Thermo). 1H (300 MHz)
and 13C (75 MHz) NMR spectra were recorded on a Bruker 300 NMR
spectrometer with a BBO probe. MS spectra were obtained on a Micro-
mass LCT high resolution time-of-flight instrument by direct injection.
Ionization was performed in positive electrospray mode. Obtaining mass
spectra of perfluorinated compounds is notoriously difficult. However,
for the homologous series perfluoralkyl NHS esters 1ꢀ6, satisfying mass
spectra were obtained for 3, 4, and 5.
1H NMR CDCl3 δ 2.90 (2H, t), 2.79 (4H, s), 2.60ꢀ2.42 (2H, m). 13
C
NMR CDCl3 δ 168.7 (CdO amide), 166.7 (CdO ester), 26.2 (CH2),
25.6 (CH2ꢀCH2), 22.8 (CH2). MS, calculated average isotopic com-
position for C15H8F17NO4, 589.0 Da; found: m/z 611.9 [M+Na].
2,5-Dioxopyrrolidine-1-yl4,4,5,5,6,6,7,7,8,8,9,9,10,11,11,11-hexade-
cafluoro-10-(trifluoromethyl)undecanoate 6. Compound 6 was pre-
pared by reaction of 2H,2H,3H,3H-perfluoro-10-methylundecanoic acid
(542 mg, 1 mmol) with TSTU (361 mg, 1.2 mmol) and DIEA (205 μL,
1.2 mmol) overnight in THF (3 mL) according to general procedure.
1
Yield 305 mg (52%). H NMR CDCl3 δ 2.99 (2H, t), 2.88 (4H, s),
2.68ꢀ2.51 (2H, m). 13C NMR CDCl3 δ 168.7 (CdO amide), 166.7
(CdO ester), 26.2 (CH2), 25.6 (CH2ꢀCH2), 22.8 (CH2).
General Procedure for Synthesis of Perfluoroalkyl Insulin
Analogues. B29Nε-2H,2H,3H,3H-Perfluorohexanoyl DesB30 Human
Insulin DesB30-C3F7. DesB30 human insulin (1, 200 mg, 35 μmol)
was dissolved in 0.1 M aq. Na2CO3 (2 mL) at pH 10.5, then active ester
2,5-dioxopyrrolidine-1-yl 4,4,5,5,6,6,6-heptafluorohexanoate (1, 16.4 mg,
42 μmol) dissolved in CH3CN (2 mL) was added. The pH was mea-
sured again to ensure a value of >10, for LysB29-selective acylation. After
45 min the reaction mixture was quenched with 0.2 M methylamine and
the crude product was precipitated by adjustment of pH to 5.5 using
0.1 M aq. HCl. Centrifugation afforded crude material which was
dissolved in water (due to solvability concern the pH was adjusted
to 8.2 using 0.1 M Na2CO3) and then purified by reverse phase HPLC
(C4 column, water/CH3CN/0.1% TFA). Insulin derivative DesB30-C3F7
was isolated by lyophilization, yielding 32.5 mg (16%). Analytical HPLC
method 1: purity >98%. Analytical HPLC method 2: purity 97%. ESI-MS,
calculated average isotopic composition for C259H379F7N64O76S6, 5930.5
Da; found: m/z 1483.5 [M+4H]4+, 1186.7 [M+5H]5+.
General Procedure for Synthesis of Succinimide Ester. 2,5-
Dioxopyrrolidine-1-yl 4,4,5,5,6,6,6-heptafluorohexanoate 1. 2H,2H,-
3H,3H-Perfluorohexanoic acid (242 mg, 1 mmol) was dissolved in
THF (3 mL), then TSTU (361 mg, 1.2 mmol) and DIEA (205 μL,
1.2 mmol) were added. The reaction mixture was stirred overnight at
room temperature, evaporated to dryness, dissolved in ethyl acetate
(50 mL), and washed with 0.1 M HCl (50 mL) and then water (50 mL).
The organic phase was dried over MgSO4 and evaporated to dryness,
B29Nε-2H,2H,3H,3H-Perfluoroheptanoyl DesB30 Human Insulin
DesB30-C4F9. Synthesis was performed according to the general proce-
dure. Analytical method 1: purity >98%, analytical method 2: purity
98%, yield 38.5 mg (18%). ESI-MS, calculated average isotopic compo-
sition for C260H379F9N64O76S6, 5980.5 Da; found: m/z 1994.4 [M+3H]3+,
1495.5 [M+4H]4+, 1196.5 [M+5H]5+.
1
yield 140 mg (41%). H NMR (CD3)2SO δ 3.0 (2H, t), 2.8 (4H, s),
2.52ꢀ2.48 (2H, m). 13C NMR CD3)2SO δ 169.7 (CdO amide), 167.7
(CdO ester), 25.4 (CH2), 25.2 (CH2ꢀCH2), 22.2 (CH2).
2,5-Dioxopyrrolidine-1-yl 4,4,5,5,6,6,7,7,7-nonafluoroheptanoate 2.
Compound 2 was prepared by reaction of 2H,2H,3H,3H-perfluorohep-
tanoic acid (292 mg, 1 mmol) with TSTU (361.3 mg, 1.2 mmol) and
DIEA (205 μL, 1.2 mmol) overnight in THF (3 mL) according to
the general procedure, yield 132 mg (34%). 1H NMR (CD3)2SO δ 3.0
(2H, t), 2.8 (4H, s), 2.51ꢀ2.48 (2H, m). 13C NMR (CD3)2SO δ 168.7
(CdO amide), 166.7 (CdO ester), 26.1 (CH2), 25.6 (CH2ꢀCH2),
22.8 (CH2).
B29Nε-2H,2H,3H,3H-Perfluoro-8-methyl-nonanoyl DesB30 Human
Insulin DesB30-C6F13. Synthesis was performed according to the general
procedure. Analytical method 1: purity >98%, analytical method 2:
purity 96%. Yield 21.5 mg (10%). ESI-MS, calculated average isotopic
composition for C262H379F13N64O76S6, 6080.6 Da; found: m/z 1520.5
[M+4H]4+, 1216.9 [M+5H]5+.
2,5-Dioxopyrrolidine-1-yl 4,4,5,5,6,6,7,7,8,9,9,9-dodecafluoro-8-(tri-
fluoromethyl)nonanoate 3. Compound 3 was prepared by reaction of
2H,2H,3H,3H-perfluoro-8-methylnonanoic acid (500 mg, 1.13 mmol)
with TSTU (409 mg, 1.36 mmol) and DIEA (232 μL, 1.36 mmol)
overnight in THF (3 mL) according to the general procedure, yield
214 mg (35%). 1H NMR CDCl3 δ 2.9 (2H, t), 2.8 (4H, s), 2.59ꢀ2.41 (2H,
m). 13C NMR (CD3)2SO δ 176.8 (CdO amide), 172.3 (CdO ester),
B29Nε-2H,2H,3H,3H-Perfluorononanoyl DesB30 Human Insulin
DesB30-C7F15. Synthesis was performed according to the general pro-
cedure. Analytical method 1: purity >98%, analytical method 2: purity
96%, yield 13.8 mg (6%). ESI-MS, calculated average isotopic composition
for C263H379F15N64O76S6, 6130.6 Da; found: m/z 1533.2 [M+4H]4+,
1226.9 [M+5H]5+.
594
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
B29Nε-2H,2H,3H,3H-Perfluoro-undecanoyl DesB30 Human Insulin
DesB30-C8F17. Synthesis was similar to compound DesB30-C3F7. Ana-
lytical method 1: purity >98%, analytical method 2: purity 96%. Yield
8.4 mg (4%). ESI-MS, calculated average isotopic composition for
p(r) function can be considered as a histogram of interparticle dis-
tances.20 Thus, both the I(q) and p(r) plots contain information about
the particle shape and size.
Modeling of the Small-Angle Scattering Data. Generally, the
scattering, I(q), from particles in solution can be calculated as a product
of form factor, P(q), describing the scattering from the single particles,
an effective structure factor, S(q), describing the interparticle scattering
and a prefactor, c, which may be calculated by the product of the
concentration of the particles, their molar mass, and their excess
scattering length per unit mass.18,21
C
264H379F17N64O76S6, 6180.6 Da, found: m/z 1546.3 [M+4H]4+,
1237.2 [M+5H]5+.
B29Nε-2H,2H,3H,3H-Perfluoro-10-methyl-undecanoyl DesB30 Human
Insulin DesB30-C9F19. Synthesis was performed according to the general
procedure. Analytical method 1: purity >97%, analytical method 2:
purity 98%. Yield 18.8 mg (8%). ESI-MS, calculated average isotopic
composition for C265H379F19N64O76S6, 6230.6 Da,found: m/z 1520.5
[M+4H]4+, 1216.9 [M+5H]5+.
IðqÞ ¼ c PðqÞ SðqÞ
ð1Þ
3
3
In order to model the form factor, direct space representations of the
different oligomeric forms of insulin were obtained from the Protein
Data Bank 3I3Z.pdb for the insulin monomer,22 while the insulin dimer
and hexamer were created by use of PyMOL23 from the 1EV3.pdb
structure. Using the Crysol24 software the scattering intensities from,
respectively, the insulin monomer, dimer, and hexamer were calculated.
These numerical representations of the scattering intensities were then
used by our Fortran based least-squares fitting routine in combination
with a structure factor describing the particleꢀparticle interaction
effects. The numerical form factor for the single proteins and the
analytical structure factor were combined using Kotlarchyk and Chen’s
so-called decoupling approximation25 using the same approach as
described in a previous article by Høiberg-Nielsen et al.26
Fractal-Like Aggregates Composed of Insulin Subunits.
One of our structural hypotheses for the aggregates formed by the
perfluoroalkyl insulin conjugates was that the perfluoroalkyl chains would
still allow the formation of hexameric insulin complexes, but the attrac-
tions between the perfluoroalkyl chains would induce formation of
oligomers of hexamers and thus of agglomerates of the insulin. A good
structural representation of the small-angle scattering of such agglomer-
ates is the fractal-like structure factor originally developed by Teixeira
et al.27 Teixeira’s structure factor for fractal aggregates was then combined
with a form factor of the single subunits of the agglomerate, in this case the
form factor for the insulin. This way the small-angle scattering of, for
example, insulin hexamers aggregating into a larger fractal-like aggregate
can be mathematically described and subsequently compared to the
experimental data. The structure factor for the fractal model is given by
eq 2; where D is the fractal dimension, r0 is the interaction radius of the
subunits in the model, Γ(x) is the gamma function, ξ is a correlation
length corresponding to the size of the aggregates.
V8 Treatment to Verify the LysB29 Acylation. DesB30-C3F7
and DesB30-C9F19 (1 mg) in phosphate buffer (0.1M, pH 7.5, 300 mL)
was treated with V8 glutamyl endopeptidase (6 mL, 1 mg/mL, 2%, w/w).
The mixture of peptide fragments was then analyzed by LC-MS, which
proved the expected acylation of LysB29.
Insulin Receptor Binding Assay. Binding of the insulin ana-
logues to the insulin receptor (IR) was determined by competition of
[125I]TyrA14-labeled insulin binding in a scintillation proximity assay
(SPA) as recently published15 and data from the SPA were analyzed
according to the four-parameter logistic model assuming a common
slope, basal and maximal response level of the curves for human insulin
and the insulin analogues. The percentage of tracer bound in the absence
of competing ligand was less than 15% in all assays, to prevent ligand-
depletion artifacts, and ∼14-fold changes in responses were obtained.
The affinities (picomolar affinity range) of the analogues are calculated
relative to that of human insulin [IC50(insulin)/IC50(analogue) ꢁ
100%] measured within the same plate.
Circular Dicroism (CD) Spectroscopy. Insulin analogues were
dissolved in 100 mM Tris buffer, pH 8.0. Concentrations were deter-
mined by UV absorbance using a molecular absorption coefficient, ε280
of 5730 Mꢀ1 cmꢀ1. Far-UV CD data were recorded on a JASCO J815
calibrated with ammonium d-10-camphorsulfonate (ACS). All spectra
were recorded at room temperature using a 0.01 cm cell path length and
a peptide concentration of about 1 mg/mL. The result Δε is based on the
molar concentration of peptide bond.
Small Angle X-ray Scattering (SAXS). Measurements were
performed on the I711 small-angle X-ray scattering beamline at MaxLab,
Lund, Sweden.16 All samples of perfluoroalkyl insulin analogues were
prepared according to a standard method for insulin sample preparation.
Insulin and perfluoroalkyl insulin analogue mixtures were dissolved
in CH3CN:H2O (1:1) and lyophilized to ensure a homogeneous
mixture. Then the samples were dissolved as in standard insulin
formulation with a total concentration of 600 μM insulin:perfluoro-
alkyl insulin, with 2 zinc(II) per hexamer (300 μM), 60 mM phenol,
100 mM NaCl, 7 mM sodium phosphate, pH 7.4. Data were collected
on a MARCCD 165 CCD detector. Buffer background measure-
ments were performed before and after each sample measurement
and then averaged before being used for background subtraction. All
SAXS measurements were performed at 24 ( 1 °C. Water was used
for the absolute calibration and checked against DesB30 insulin hexa-
mers (∼3 mg/mL).
1
DΓðD ꢀ 1Þ
SðqÞ ¼ 1 þ
h
iðD ꢀ 1Þ=2
D
ðqr0Þ
1 þ 1=ðq2ξ2Þ
ꢀ
ꢁ
ꢁ sin ðD ꢀ 1Þtanꢀ1ðqξÞ
ð2Þ
In the case of weak attraction between the modified insulin analogues,
the aggregates will assemble and disassemble in a dynamical fashion and
both hexamer aggregates and “monomeric” hexamers will be present in
the sample. This situation is taken into account in the modeling by
allowing for a microstructure where a fraction, XSA, of the insulin is in the
form of self-assembled fractal-like aggregates of hexamers and the
remaining fraction, (Xhex = 1 ꢀ XSA), is in the form of free hexamers.
Particles with Cylindrical Geometry ꢀ Short Cylinders. The
first fractal model evaluated could not adequately describe the scattering
from the insulin with the C7F15 and C8F17chains. Instead the DesB30-
C7F15 and DesB30-C8F17 insulins were best fitted with a model for
particles with local cylindrical geometry. For the DesB30-C7F15 we used
the form factor short cylindrical particles.28,29
Using the BioXTAS RAW17 software, the 2-D CCD images were
transformed into the 1-D I(q) representation where q = 4π sin θ/λ,
where λ is the X-ray wavelength (λ = 1.1 Å in the present experiment)
and θ is the scattering angle.18 Furthermore, the data were background
subtracted and converted into absolute scale units (cm2/cm3) by using
water as an external standard. In the process of preliminary data analysis,
the obtained scattering intensities, I(q), were converted into a direct
space representation in terms of the pair distance distribution function,
p(r), by means of indirect Fourier transformation using the IFT method
based on Bayesian statistics that has been built into BioXTAS RAW.
For proteins, which have an almost uniform electron density,19 the
"
#
2
Z
2J1ðqR sin αÞ sinðqL cos α=2Þ
ð3Þ
PcylinderðqÞ ¼
sin α dα
qR sin α
qL cos α=2
595
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
Scheme 1a
a (A) Synthesis of succinimide esters (1ꢀ6). (B) Synthesis of insulin analogues with linear and branched perfluoroalkyl chains at the LysB29 side-chain.
Where J1(x) is the first-order Bessel function, R is the cross-section
radius, and L is the length of the cylinder.
Particles with Cylindrical Geometry ꢀ Wormlike Chains.
For DesB30-C8F17 the best model fits, somewhat surprisingly, were
obtained with a model for wormlike chains with local cylindrical geo-
metry. The scattering from the semiflexible self-avoiding wormlike
chains may be factorized into a term describing the chain cross-section
contribution to the scattering, PCS(q), and a term PWC(q), describing the
longitudinal contribution to the scattering in terms of a semiflexible
wormlike chain.30,31
PðqÞ ¼ PCS ꢁ PWCðqÞ
ð4Þ
The contribution to the scattering from the local cylindrical geo-
metry is
"
#
2
2 J1ðqRÞ
3
PCSðqÞ
¼
ð5Þ
qR
Figure 1. Far-UV CD of the six analogues.
Where J1(x) is the first-order Bessel function and R is cross-section
radius.
The longitudinal contribution to the scattering, PWC(q), was calcu-
lated using a previously applied30 numerical expression for the scattering
from semiflexible wormlike chains as developed by Pedersen and
Schurtenberger.31 The expression is based on a series of Monte Carlo
simulations of semiflexible polymers with excluded volume effects,
performed by Pedersen, Laso, and Schurtenberger.32 The authors
simulated a wide range of polymer configurations, sampled them, and
calculated and parametrized the corresponding scattering functions,
thus obtaining expressions for the scattering functions of semiflexible
wormlike micelles in terms of the chain contour length L and the local
rigidity as represented by the Kuhn length b.
8.4 and 7.1 for GlyA1 and PheB1, respectively, and 11.2 for the
ε-amino group of LysB29. As a result, the NHS-esters 1ꢀ6 can
selectively react at LysB29 at high pH (∼10) to form an amide
linkage (Scheme 1B).34 The perfluoroalkyl B29-derivatives were
the main products and were purified to homogeneity and
analyzed under two different HPLC conditions (see Supporting
Information Table S1 for purity and HPLC conditions).
To verify that only LysB29 had been acylated, a V8 glutamyl
endopeptidase cleavage was performed on compounds DesB30-
C3F7 and DesB30-C9F19 followed by HPLC-MS analysis of the
fragments, which clearly showed that only the expected LysB29
conjugates were obtained.35
CD Spectroscopy. CD spectroscopy was used to study the
secondary structure for any conformational changes after con-
jugation. The far-UV CD band at 208 nm primarily arises from an
α-helical structure, as does the band at 222 nm.36 The spectral
characteristics of the perfluorinated insulin analogues with short-
er chain length (DesB30-C3F7 and DesB30-C4F9) were found to
be very similar to desB30 insulin. However, CD spectra of insulin
The structure factor for the particles with cylindrical geometry was
assumed to be unity, thus reducing eq 1 to I(q) =c P(q).
3
’ RESULTS
Chemical Synthesis. The succinimidyl esters of six different
perfluoroalkyl chains were synthesized and conjugated to the
ε-amino group of LysB29 DesB30 insulin (Scheme 1A).33 The
pKa values of the two N-terminal α-amino groups of insulin are
596
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
Figure 2. (A) SAXS data of the different perfluoroalkyl-insulin analogues. (B) the corresponding pair distance distribution (p(r)-functions). (C)
The p(r)-functions plotted in the region 0ꢀ120 Å.
conjugate DesB30-C9F19 which carries the longest perfluoroalkyl
chain, revealed that the substitution had disrupted part of the
helical structure of the molecule (Figure 1).
SAXS Measurements. SAXS was used to examine the self-
assembly of the insulin analogues which were formulated as
above with ZnII and phenol. Hexamer DesB30 insulin was used
as a standard. The experimental scattering profiles for DesB30-
C3F7 and DesB30-C4F9 look remarkably similar to insulin
DesB30 itself (Figure 2A). This strongly suggests that the
hexameric structure of the DesB30 is conserved in these two
samples and that the attachment of short perfluoroalkyl chains
does not significantly affect the local structure. For the DesB30-
C6F13 the high-q part of the I(q) resembles that of the DesB30,
while the low-q part has an increased scattering intensity. This
indicates the formation of larger aggregates build from hexameric
subunits. For DesB30-C7F15, DesB30-C8F17, and DesB30-
C9F19, the scattering intensity of the low-q region increases
even further and a clear trend of an increasing fraction of self-
assembled hexamers with increasing flourine-content can be seen
for the perfluoroalkyl insulins. I(q) data for DesB30-C7F15,
DesB30-C8F17, and DesB30-C9F19 furthermore shows that the
Receptor Binding. The insulin receptor affinities of the six
perfluoroalkyl insulin analogues were tested as previously des-
cribed,37 and the affinities for analogues DesB30-C3F7 and Des
B30-C4F9 were comparable to insulin (93% and 77%, respectively),
whereas analogues DesB30-C6F13 and DesB30-C7F15 gave 2-fold
reductions in binding affinity, 54% and 46%, respectively (see
Supporting Information Table S2 for insulin receptor binding
affinity). DesB30-C8F17 and DesB30-C9F19 had a binding affinity
of 9.1% and 6.2%, respectively, compared to human insulin. Thus,
when the size of the perfluoroalkyl substituent was increased, the
binding affinity decreases in this case. However, a ∼10% relative
binding still corresponds to an affinity (IC50) in the picomolar range
for the insulin receptor and full in vivo potency of insulin analogues
has been observed despite lowered in vitro insulin receptor
affinities.38
597
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
Figure 3. SAXS data and model fits. (A) Des-B30 insulin fitted with the model for insulin hexamers. (B) DesB30-C3F7, (C) DesB30-C4F9, and
(D) DesB30-C6F13 are all fitted with the model for an insulin hexamer as the basic unit combined with a structure factor for fractal aggregates
of insulin hexamers. (E) DesB30-C7F15 fitted with model for short cylinders. (F) DesB30-C8F17 fitted with a model for Gaussian chains with
local cylindrical geometry.
high-q region gradually deviates more and more from that of the
pure DesB30 insulin. This indicates that the larger aggregates are
no longer composed of the hexameric subunits.
Structural Modeling of the Aggregates. The above-described
visual inspection of the experimental scattering functions and the
derived pair-distance distribution functions indicate that the
aggregates formed by the analogues with short perfluoroalkyl
chains have hexamers as their building blocks, while another and
probably less ordered microstructure is predominant in the
aggregates with long perfluoroalkyl chains. A series of structurally
based simulations and corresponding model fits were performed
to test these conclusions in a more quantitative way. For the
structural modeling we assumed that the insulin was in the form of
monomers, dimers, or hexamers and that these would then con-
stitute the basic building blocks of larger aggregates. We further-
more assumed that the formed larger aggregates were randomly
structured agglomerates of the subunits. As described above, such
structures have previously been successfully described by the so-
called Fractal structure factor model developed by Teixeira et al.27
In addition to these intuitively expected structures we tried to fit
some of the data with form factors corresponding to polydisperse
spheres, discs, and rigid and flexible rods.
This set of observations is also reflected in the pair distance
distribution functions (p(r)-functions) obtained from the indir-
ect Fourier transformed data (IFT) (Figure 2B).20 Here the
nearly bell shaped p(r)-functions for DesB30 insulin and analo-
gues DesB30-C3F7 and DesB30-C4F9 look very similar, with a
Dmax of approximately 50 Å, and indicate a compact and close to
spherical structure as is seen for hexameric insulin. The p(r)-
function for analogue DesB30-C6F13 is still bell-shaped at low
r values but with a significant tail at high r values indicating the
formation of larger assemblies with the hexamer as the subunit, as
illustrated in Figure 5. The p(r)-function for analogue DesB30-
C7F15 clearly differs from that of the desB30 insulin, revealing the
formation of even larger molecular self-assemblies.
Taken together, the visual inspection of scattering profiles and
the p(r)-functions for the six perfluoro-insulin analogues demon-
strated that the sizes of the assemblies increase with the number
of fluorine atoms in the perfluoroalkyl chains, which can be seen
from increasing I(0) or larger p(r) curve area.
Rewardingly, it was possible to fit the three data sets with atta-
ched short perfluoroalkyl chains (DesB30-C3F7, DesB30-C4F9,
598
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
Table 1. Model Results for the Short Perfluoralkyl Insulinsa
Table 2. Model Results for the Long Perfluoroalkyl Chain
Insulinsa
ID
XSA/%
XHex/%
ξ [Å]
Dfrac
ID
Rchain [Å]
L [Å]
bKuhn [Å]
DesB30
0
100
70
n.a.
n.a.
DesB30-C7F15
DesB30-C8F17
20.9 ( 0.1b
20.2 ( 0.6c
66.1 ( 0.2b
3600 ( 1000c
n.a.
22.5 ( 8.4c
DesB30-C3F7
DesB30-C4F9
DesB30-C6F13
30 ( 4
9 ( 1
59 ( 5
42 ( 11
42b
42b
1.13 ( 0.33
1.13b
91
a Self-assembly into cylindrical and worm-like structures. b Short cylin-
41
1.06 ( 0.04
drical rods. c Worm-like chains with cylindrical cross-section. Rchain
Cross-sectional radius of the cylinder. L: contour length of the cylinder.
Kuhn: The Kuhn length of the worm-like chains. d Parameter set
:
a Hexamer subunits organized in fractal-like aggregates. XSA: Fraction of
hexamers self-assembled into a fractal-like structure. XHex: Fraction of
non-self-assembled hexamers calculated as (1 ꢀ XSA). ξ: The correlation
length of the aggregates. Dfrac: The fractal dimension. b Parameter set
manually.
b
manually.
wormlike chains) were tried. The best agreement between model
and experimental data was obtained when using a model for
insulin aggregates having, respectively, cylindrical and wormlike
structures. In the case of DesB30-C7F15 a cylinder radius slightly
larger than a dimer but smaller than a hexamer was found. As can
be seen from Figure 3E DesB30-C7F15 has a Guinier region,
which fits well with that the model fit gives a well-determined
average cylinder length, L, of no more than 66 Å (Table 2). The
DesB30-C8F17 particles have approximately the same radius as
the DesB30-C7F15 particles, but are much longer. The attractions
introduced by the C8F17-appendix lead to the assembly of very
elongated wormlike chains. A good model-fit could be obtained
with a chain-length of about 3600 and a Kuhn length of 22.5 Å
(Table 2). Mathematically, there was a strong internal correlation
between the chain length and the Kuhn length in the model-fits;
this gave rise to relatively large error bars on the obtained values
for these two parameters. However, the average radius of the
chain could be determined with great accuracy.
Interpretation of the SAXS data suggested a structural be-
havior of the different insulin analogues as a function of the
perfluoroalkyl chain length as is illustrated in Figure 5. Insulins
with short perfluoroalkyl chains have a hexameric structure
similar to human insulin. As the length of the perfluoroalkyl
chains increases, attraction between the hexamers is introduced
leading to a fractal-like structure with the hexamer as the
fundamental building block. With a further increase in the length
of the perfluoroalkyl chain, the fluorous interactions increase
further and surpass the interactions that lead to hexamer for-
mation. This leads to a breakdown of the hexameric substruc-
ture and the appearance of a wormlike structure. How can the
formation of such a chainlike structure with subunits smaller than
the hexamer be explained? A simple explanation would be that
the chain is held together by alternation between the insulin
dimer interactions and the fluorous interactions. This hypothe-
tical structure is illustrated in Figure 5.
and DesB30-C6F13) with a model where we assumed that a
fraction of the hexamers were organized in a fractal-like structure,
while the remaining part was still in the form of free hexamers
(Figure 3A,B,C,D). However, for the particles with the longest
perfluoroalkyl chains (DesB30-C7F15, DesB30-C8F17, and Des
B30-C9F19) all our attempts to describe the data as resulting from
aggregates composed from insulin subunits broke down, when
we assumed insulin monomers, dimers, and hexamers as the basic
building blocks. The most likely explanation is that the oligo-
meric form of the insulin is no longer well-defined due to the very
strong attraction between the perfluoroalkyl chains. Instead we
sought to get a picture of the overall geometry of the formed
aggregates by fitting the form factors corresponding to the above-
mentioned different-shaped particles to the data. Interestingly,
for DesB30-C7F15 and DesB30-C8F17 we found a good agree-
ment between the experimental data and, respectively, a model
for short cylinders and a model for Gaussian chains with local
cylindrical geometry (Figure 3E,F). We were not able to obtain
satisfactory fits to the scattering data from the DesB30-C9F19
sample, i.e., the sample with the longest attached perfluoroalkyl
chain, with any of the models.
Samples with Insulin Self-Assembled into a Fractal Like
Structure. DesB30-C6F13 was fitted with the fractal-like model
(Table 1), with the fraction of self-assembled hexamers (XSA) fixed
to 1. This gave a correlation length, ξ, around 42 Å and a fractal
dimension, Dfrac, of 1.06. Thus a fractal structure with a correlation
length of the approximate size of the insulin hexamer and a slightly
branched structure. DesB30-C3F7 and DesB30-C4F9, both exhib-
ited a lower fraction of self-assembly than DesB30-C6F13. In order
to avoid overparametrizing the data, Dfrac was fixed to the value
obtained for DesB30-C6F13 suchthatonly thefraction ofhexamers
self-assembled into a fractal-like structure, XSA, and the correlation
length, ξ, were fitted. For the DesB30-C4F9, ξ, also had to be fixed
in order to limit the degree of freedom in the model fits and was set
to the value found for DesB30-C6F13.
Self-Assembly by Formation of Mixed Aggregates of Per-
fluoroalkyl Modified and Nonmodified Insulin. Based on these
observations we studied whether an additional level of control of
self-assembly could be introduced by forming hexamers consist-
ing of a mixture of fluoroalkyl insulins and unmodified insulin
such that hexamers would have fewer fluorous contact points.
This approach would potentially give a distribution of different
hexamers.
For DesB30-C3F7, ξ was found to be 42 Å, which is comparable
to, however, slightly smaller than the size of the insulin hexamer.
XSA was found to be 30% givinga smaller fraction of self-assembled
hexamers compared to DesB30-C6F13. With a XSA of 9% DesB30-
C4F9 is, somewhat surprisingly, the perfluoroalkyl insulin investi-
gated with the lowest fraction of self-assembled hexamers.
Using this approach, the possibility of controlling the formation
of large assemblies formed by DesB30-C9F19 insulin was investi-
gated. A series of formulations of a mixture of unmodified insulin
and analogue DesB30-C9F19 were prepared with ZnII and phenol,
such that the ratio was always optimal for hexamer formation
regarding ZnII and phenol. The SAXS data and the corresponding
p(r)-functions obtained are shown in Figure 4A and B.
Samples with Insulin Aggregated into a Local Cylindrical
Structure. The SAXS data from the DesB30-C7F15 and DesB30-
C8F17 insulins could not be fitted with the model for fractal-like
aggregates: neither when the hexameric subunits were applied,
as for the short perfluoroalkyl insulins, nor when the insulin
monomer or dimers were used as the subunits. Instead different
simple geometrical models (spheres, ellipsoidal discs, cylinders,
599
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
Figure 4. (A) SAXS data from the mixtures of unmodified DesB30 insulin and perfluoroalkyl insulin DesB30-C9F19 with IFT fit. (B) The corresponding
p(r)-functions. (C) SAXS data for the DesB30-C9F19:DesB30 insulin (2:4) mixture, solid line: experimental data, broken line the superposition of the
SAXS data obtained from pure DesB30 and DesB30-C9F19 combined. The fact that the experimental data for the mixture differs from the combined data
for DesB30 and DesB30-C9F19 indicates that the DesB30 and DesB30-C9F19 mixture does not self-segregate but forms a mixture at the molecular level.
In both the I(q) functions and in the derived p(r)-functions it
was clearly seen that the average size of the aggregates increased
with higher proportion of fluoro-insulin analogue DesB30-C9F19
in the mixture (Figure 4). The absence of a well-defined Guinier
range in the I(q) data, however, made it difficult to accurately
quantify the maximal size of the formed aggregates. Nevertheless,
our data clearly demonstrated that the formation of large assem-
blies formed by analogue DesB30-C9F19 can to some extent be
controlled by dilution with unmodified desB30 insulin to provide
hexamers with <1 fluorous contact point per hexagonal face.
However, in the case of the very strong attractions induced by
the C9F19 chains, this dilution was not sufficient to reintroduce
an ordered hierarchical structure based on, for example, the hexa-
meric insulin structure as the main building block.
In this case the SAXS data of a mixed sample should simply be a
superposition of the SAXS data from the unmodified and perfluoro-
alkyl insulin analogues weighted by their relative concentrations.
That this is certainly not the case is clearly shown from Figure 3C.
The data for the DesB30-C9F19:DesB30 insulin (2:4) mixture
(Figure 4C blue line) was compared to an appropriately weighted
superposition of the SAXS data obtained from the two pure
mixtures, thus pure DesB30 (2/6 part) and pure DesB30-C9F19
(4/6 part) (Figure 4 blue dashed line). As is clearly seen in Figure 4C
neither the low-q nor the high-q part of the data curves agree with
the superpositioned data. We take this as a clear indication that
the two types of insulin do actually form mixed aggregates.
’ DISCUSSION
It might be speculated that the strong attraction introduced by the
perfluoroalkyl chains in insulin DesB30-C9F19 could lead to a total
segregation between the unmodified and perfluoroalkyl insulin.
Insulin readily forms dimers as well as well-defined hexamers
by complexation with two ZnII ions, which can occur in two
600
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
Figure 5. Illustration of perfluoroalkyl insulin self-assembly. Bottom: DesB30 insulin. Middle: DesB30 insulin with medium length perfluoroalkyl
chains. Top: Insulin with long perfluoroalkyl chains assembling into cylindrical structures of insulin dimers.
states. The oligomeric state of new insulin variants, i.e., whether
they are preferably monomer/dimer or hexamer or multimers of
hexamers, determines whether they are fast-acting or slow-acting
in the treatment of diabetes. Several insulin variants in clinical use
contain a prosthetic group at Lys B29, for example, lipid moieties
for binding to human serum albumin and to promote formation
of higher aggregates. Insulin is thus a highly interesting model
protein for the evaluation of new principles to control inter-
molecular self-assembly. The C-terminal Thr of the B-chain in
insulin (ThrB30) is not essential for binding, and human insulin
desB30 has been used in the clinic. The tendency of highly
fluorinated molecules to avoid contact with hydrophilic and
hydrophobic surfaces and to prefer contact with fluorinated
groups leads to a self-segregation which is often referred to as
the fluorous effect. This has been used for the isolation and
handling of small molecules, as well as to direct the intramole-
cular folding of peptides and proteins.14 In contrast, in this
project we incorporated perfluoroalkyl moieties on the surface of
a small protein, insulin, to direct and control the intermolecular
self-assembly, hence the quarternary structure. Here, a series of
insulin variants functionalized with perfluoroalkyl chains, C3F7 to
C9F19, were synthesized by pH controlled regioselective acyla-
tion of the Lys side-chain of B29 of desB30 human insulin. These
six new insulin variants were then analyzed for their insulin
receptor binding, their secondary structure, and their self-assem-
bly into larger structures.
Insulin variants carrying the perfluoroalkyl chains C3F7, C4F9,
C6F13, and C7F15 had insulin receptor affinities of 93ꢀ46%
compared to human insulin. This is still in the picomolar range
and the somewhat lowered affinities are not detrimental. The
analogues carrying C8F17 and C9F19 chains had affinities of 9.1%
and 6.2%, respectively, compared to human insulin. Thus, per-
fluoroalkyl modification can be compatible with maintaining
receptor binding. These values are also comparable to a recently
published paper on metal ion mediated self-assembly of insulin
derivatives carrying a bipyridine ligand.39
A comparison of the CD spectra of the six perfluoralkyl
DesB30 insulin variants indicated that only the insulin carrying
the longest perfluoroalkyl chain, C9F19, had a significantly
lowered degree of α-helicity. The other variants had helicities
similar to human insulin or only slightly lower, which was
encouraging.
Analysis of the SAXS data of the insulin variants carrying short
perfluoroalkyl chains (C3F7, C4F9, and C6F13) revealed that the
sizes of the assemblies increased with the number of fluorine
atoms in the perfluoroalkyl chains. A more detailed structural
modeling based data analysis further suggested that the hexa-
meric subunits were conserved as basic building blocks up to a
601
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
perfluoroalkyl chain length of C6F13. However, with longer per-
fluoroalkyl chain length, a breakdown of the hexameric sub-
structure was observed and instead a new structure, apparently
with local cylindrical geometry, became dominating, as illustrated
in Figure 5. This new cylindrical geometry was observed for
DesB30-C7F15 and DesB30-C8F17 and we observed that the
increased molecular attraction introduced by the longer per-
fluoroalkyl chains also gave rise to a significant increase in the
overall length of the cylindrical structures. Thus, it seems that
with short perfluoroalkyl appendices the hexameric subunit core
is still conserved as the dominant structure, whereas with longer
perfluoroalkyl appendices, it breaks down and the fluorous
interactions start to override and promote intermolecular self-
assembly leading to larger structures in combination with a
breakdown of the local hexameric structure.
We were not able to make any satisfactory fits to the scattering
data from the DesB30-C9F19 sample with any of the models. A
possible explanation could be that the sample is composed of
differenttypes ofinsulinsubunits andthatthe overall structure, asa
result of the very strong interaction introduced by the C9F19,
becomes disordered. This again indicates that the very strong
fluorous interaction overrules the formation of hexamer constructs
leading to the formation of larger aggregates. Thus, the C9F19
chain may not be useful for controlling protein self-assembly, if it is
used as a uniform modification. However, a different approach for
using the C9F19 modification was designed, where unmodified
human insulin desB30 was mixed with human insulin DesB30-
C9F19 in different ratios to study whether structural control could
be regained by molecular dilution of the DesB30-C9F19 with
unmodified desB30. Indeed, SAXS data indicated that the un-
modified insulin and insulin formed mixed oligomers; however,
this molecular dilution was, according to our interpretation, not
sufficient to reintroduce more well-ordered structures.
’ CONCLUSIONS
We have prepared six new insulin variants carrying perfluor-
oalkyl moieties by regioselective chemical acylation of LysB29 in
desB30 insulin at high pH. The perfluoroalkyl chain lengths were
varied systematically so that the level of fluorous interactions
between the perfluoralkyl insulins was controlled. The six new
insulin variants maintained satisfactory levels of binding to the
insulin receptor. All but one of them maintained a high degree of
α-helicity, as in human insulin. The self-assembly of the per-
fluoroalkyl insulin variants was studied by SAXS. We introduced
a new and bio-orthogonal concept in intermolecular protein self-
assembly, which relies on the self-segregating behavior of per-
fluoroalkyl moieties. Hierarchical intermolecular self-assembly
was achieved by combining classical insulin hexamer formation
by addition of ZnII and phenol, with incorporation of perfluor-
alkyl-insulins with different perfluoroalkyl chain lengths. Re-
wardingly, three different types of self-assembly were achieved
depending on the size of the perfluoroalkyl group. Relatively
short perfluoroalkyl chains did not influence the overall quar-
ternary structure of insulin, whereas the longer perfluoroalkyl
chains provided increased degrees of self-association of insulin
hexamers. The formation of mixed hexamers of insulins and
insulins carrying longer perfluoroalkyl chains provided a further
level of control. We envision that this approach can be adapted to
other proteins and lead to a new way to control self-assembly of
proteins.
’ ASSOCIATED CONTENT
S
Supporting Information. Additional information as de-
b
scribed in the text. This material is available free of charge via the
Internet at http://pubs.acs.org.
The perfluoroalkyl moiety serves a different purpose than a
lipid chain on a protein, as the latter promotes binding to hy-
drophobic surfaces such as serum albumin, while the former
uses the fluorous interactions to promote intermolecular self-
assembly. These results indicate that modification of proteins
with a perfluoroalkyl moiety can indeed be used to intermole-
cular assemble them. Importantly, the size of the perfluoroalkyl
group provides a level of control. Longer perfluoroalkyl chains
could in the present study override metal ion mediated self-
assembly and gave cylindrical structures composed of insulin
dimers, whereas medium-length perfluoroalkyl chains gave frac-
tal-like assemblies composed of insulin hexamers. This is illu-
strated in Figure 5 where we go from DesB30 insulin assembly
into hexamers (Figure 5 bottom), and then by tuning the length
of the perfluoroalkyl chain, we can create fractal-like structures of
hexamers (Figure 5 middle). With further elongation of the
perfluoroalkyl chain length we can control the self-assembly into
cylindrical structures of insulin dimers (Figure 5 top). This bio-
orthogonal approach thus provides a tunable tool for construc-
tion of protein superstructures.
Using perfluoroalkyl appendices with different lengths, we
were able to establish several different scenarios for control of
self-assembly of human insulin, as illustrated in Figure 5. The
fluorous self-assembly was combined with native ZnII mediated
self-assembly, and amounts to a hierarchical process. As this
fluorous strategy is bioorthogonal, it seems likely that it can be
transferred to other proteins, for example, in the development
of biopharmaceutical drug candidates or for protein material
science.
’ AUTHOR INFORMATION
Corresponding Author
*E-mail: kjj@life.ku.dk (K.J.J); lia@life.ku.dk (L.A.); tshj@
novonordisk.com (T.H.-J).
’ ACKNOWLEDGMENT
A NABIIT grant from DSF is gratefully acknowledged.
Dr. Thomas B. Kjeldsen is thanked for insulin receptor affinity
assays. Beamtime at the I711 beamline at the MAX-lab synchro-
tron along with user support from Dr. T. Plivelic are gratefully
acknowledged. Jesper C. Westergaard is acknowledged for help
with Figure 5.
’ REFERENCES
(1) Jorgensen, L.; Hostrup, S.; Moeller, E. V.; Grohganz, H. Expert
Opin. Drug Delivery 2009, 6, 12119–11230.
(2) Owens, D. R. Nat. Rev. Drug Discovery 2002, 1, 529–540.
(3) Cefalu, W. T. Diabetes Care 2004, 27, 239–246.
(4) Baker, E. N.; Blundell, T. L.; Cutfield, J. F.; Cutfield, S. M.;
Dodson, E. J.; Dodson, G. G.; Hodgkin, D. M. C.; Hubbard, R. E.; Isaacs,
N. W.; Reynolds, C. D.; Sakabe, K.; Sakabe, N.; Vijayan, N. M. Philos.
Trans. R. Soc., B 1988, 319, 369–456.
(5) Hermansen, K.; Fontaine, P.; Kukolja, K. K.; Peterkova, V.; Leth,
G.; Gall, M.-A. Diabetologia 2004, 47, 622–629.
(6) Kaarsholm, N. C.; Ko, H. C.; Dunn, M. F. Biochemistry 1989,
28, 4427–4435.
(7) Smith, G. D.; Dodson, G. G. Proteins: Struct., Funct., Genet. 1992,
14, 401–408.
602
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603

Langmuir
ARTICLE
(8) Derewenda, U.; Derewenda, Z.; Dodson, E. J.; Dodson, G. G.;
Reynolds, C. D.; Smith, G. D.; Sparks, C.; Swenson, D. Nature 1989,
338, 594–596.
(9) Smith, G. D.; Ciszak, E.; Magrum, L. A.; Pangborn, W. A.;
Blessing, R. H. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2000.
(10) Markussen, J.; Havelund, S.; Kurtzhais, P.; Andersen, A. S.;
Halstrom, J.; Hasselager, E.; Larsen, U. D.; Ribel, U.; L. Schiiffer, K. V.;
Jonassen, I. Diabetologia 1996, 39, 281–288.
(11) Heise, T.; Nosek, L.; Rønn, B. B.; Endahl, L.; Heinemann, L.;
Kapitza, C.; Draeger, E. Diabetes 2004, 53, 1614–1620.
(12) (a) Jonassen, I.; Havelund, S.; Ribel, U.; Plum, A.; Loftager, M.;
Hoeg-Jensen, T.; Volund, A.; Markussen, J. Pharm. Res. 2006, 23, 49–55.
(b) Jonassen, I.; Havelund, S.; Ribel, U.; Hoeg-Jensen, T.; Steensgaard,
D. B.; Johansen, T.; Haahr, H.; Nishimura, E.; Kurtzhals, P. Diabetes
2010, 59 (S1), A11.
(13) Gladysz, J. A.; Curran, D. P.; Horvꢀath, I. T. Handbook of
Fluorous Chemistry; Wiley-VCH: Weinheim, 2004.
(14) (a) Gottler, L. M.; de la Salud Bea, R.; Shelburne, C. E.;
Ramamoorthy, A.; Marsh, E. N. G. Biochemistry 2008, 47, 9243–9250.
(b) Tang, Y.; Ghirlanda, G.; Petka, W. A.; Nakajima, T.; DeGrado, W. F.;
Tirrell, D. A. Angew. Chem. Int. Ed., 2001, 40, 1494–1496.
(15) Glendorf, T.; Stidsen, C. E.; Norrman, M.; Nishimura, E.;
Sørensen, A. R.; Kjeldsen, T. PLoS One 2011, 6, e20288.
(16) Knaapila, M.; Svensson, C.; Barauskas, J.; Zackrisson, M.;
Nielsen, S. S.; Toft, K. N.; Vestergaard, B.; Arleth, L.; Olsson, U.;
Pedersen, J. S.; Cerenius, Y. J. Synchrotron Radiat. 2009, 16, 498–504.
(17) Nielsen, S. S.; Toft, K. N.; Snakenborg, D.; Jeppesen, M. G.;
Jakobsen, J. K.; Vestergaard, B.; Kutter, J. P.; Arleth, L. J. Appl. Crystal-
logr. 2009, 42, 959–964.
(18) Glatter, O.; Kratky, O., Eds. Small Angle X-ray Scattering;
Academic Press: London, 1982.
(19) Mylonas, E.; Svergun, D. I. J. Appl. Crystallogr. 2007, 40, 245–
249.
(20) Glatter, O. J. Appl. Crystallogr. 1977, 10, 415–421.
(21) Pedersen, S. J. Modelling of Small-Angle Scattering Data from
Colloids and Polymer Systems; in Neutrons, X-rays and Light, Scattering
Methods applied to Soft Condensed, Matter, P., Lindner, Th. Zemb, Eds.;
Elsevier: North-Holland, 2002.
(22) www.pdb.org, 3I3Z.pdb 2010-05-10.
(23) Delano, W. L. The PyMOL molecular graphics system, version
1.2rl; Schr€odinger LLC, 2009.
(24) Svergun, D., I.; Barberato, C.; Koch, M., H, J. J. Appl. Crystallogr.
1995, 28, 768–773.
(25) Kotlarchyk, M.; Chen, S.-H. J. Chem. Phys. 1983, 79, 2461–
2469.
(26) Høiberg-Nielsen, R.; West, P.; Skov, L., K.; Arleth, L. Biophys. J.
2009, 97, 1445–1453.
(27) Teixeira, J. J. J. Appl. Crystallogr. 1988, 21, 7781–7785.
(28) Mittelbach, P.; Porod, G. Acta Physica Austriaca 1961,
14, 404–439.
(29) Fournet, G. Bull. Soc. Fr. Mineral. Crist. 1951, 74, 39–113.
(30) Arleth, L.; Bergstr€om, M.; Pedersen, J. S. Langmuir 2002,
18, 5343–5353.
(31) Pedersen, J. S.; Schurtenberger, P. Macromolecules 1996, 29.
(32) Pedersen, J. S.; Laso, M.; Schurtenberger, P. Phys. Rev. E 1996,
54, R5917–R5920.
(33) Bannwarth, W.; Knorr, R. Tetrahedron Lett. 1991, 32, 1157–1160.
(34) Gao, J.; Mrksich, M.; Gomez, F. A.; Whitesides, G. M. Anal.
Chem. 1995, 67, 3093–3100.
(35) Tofteng, A. P.; Jensen, K. J.; Schaeffer, L.; Hoeg-Jensen, T.
ChemBioChem 2008, 9, 2989–2996.
(36) Pocker, Y.; Biswas, S. B. Biochemistry 1980, 19, 5043–5049.
(37) Glendorf, T.; Soerensen, A. R.; Nishimura, E.; Pettersson, I.;
Kjeldsen, T. Biochemistry 2008, 47, 4743–4751.
(38) Ribel, U.; Hougaard, P.; Drejer, K. S., A. R. Diabetes 1990,
39, 1033–1039. Diabetes 1990, 39, 1033–1039.
(39) Munch, H. K.; Heide, S. T.; Christensen, N. J.; Hoeg-Jensen, T.;
Thulstrup, P. W.; Jensen, K. J. Chem.—Eur. J. 2011, 17, 7198–7204.
603
dx.doi.org/10.1021/la203042c |Langmuir 2012, 28, 593–603