4S
-Hydroxylation of insulin at ProB28 accelerates hexamer
dissociation and delays fibrillation
Seth A. Lieblich
†,‡
,
Katharine Y. Fang
†,‡
,
Jackson K. B. Cahn
†
,
Jeffrey Rawson
§,
‖
,
Jeanne
LeBon
§
,
H. Teresa Ku
§,
‖
,¶
, and
David A. Tirrell
†,*
†
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena,
CA 91125, USA
§
Department of Translational Research and Cellular Therapeutics, Diabetes and Metabolism
Research Institute, City of Hope, Duarte, CA 91010, USA
‖
Beckman Research Institute of City of Hope, Duarte, CA 91010, USA
¶
Irell & Manella Graduate School of Biological Sciences, City of Hope, Duarte, CA 91010, USA
Abstract
Daily injections of insulin provide lifesaving benefits to millions of diabetics. But currently
available prandial insulins are suboptimal: The onset of action is delayed by slow dissociation of
the insulin hexamer in the subcutaneous space, and insulin forms amyloid fibrils upon storage in
solution. Here we show, through the use of non-canonical amino acid mutagenesis, that
replacement of the proline residue at position 28 of the insulin B-chain (ProB28) by (4
S
)-
hydroxyproline (Hzp) yields an active form of insulin that dissociates more rapidly, and fibrillates
more slowly, than the wild-type protein. Crystal structures of dimeric and hexameric insulin
preparations suggest that a hydrogen bond between the hydroxyl group of Hzp and a backbone
amide carbonyl positioned across the dimer interface may be responsible for the altered behavior.
The effects of hydroxylation are stereospecific; replacement of ProB28 by (4
R
)-hydroxyproline
(Hyp) causes little change in the rates of fibrillation and hexamer disassociation. These results
demonstrate a new approach that fuses the concepts of medicinal chemistry and protein design,
and paves the way to further engineering of insulin and other therapeutic proteins.
Blood glucose levels are tightly controlled in mammals through a sensitive regulatory
system mediated by insulin, a 51-amino acid endocrine hormone composed of two disulfide-
linked polypeptide chains (designated A and B). Upon binding to its receptor, insulin
initiates a signaling cascade that accelerates glucose uptake and glycogen production. In
diabetic patients, this system malfunctions, and glucose levels must be controlled through
subcutaneous injections of insulin
1
. The
C
-terminus of the B-chain is important in mediating
dimerization of the hormone
2
–
3
, and the flexibility of the B-chain
C
-terminus is believed to
*
Corresponding author. tirrell@caltech.edu.
‡
Authors contributed equally to this work.
Supporting Information
The following files are available free of charge via the internet at
http://pubs.acs.org
. Experimental procedures and methods,
supporting figures and tables, additional references.
HHS Public Access
Author manuscript
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Published in final edited form as:
J Am Chem Soc
. 2017 June 28; 139(25): 8384–8387. doi:10.1021/jacs.7b00794.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
contribute to aggregation through formation of amyloid fibrils
4
–
6
. Pharmaceutical
formulations of insulin are stabilized with respect to fibrillation by addition of zinc and
phenolic preservatives, which drive assembly of the R
6
hexamer (Fig. 1a)
7
–
9
. The R
6
form of
insulin is inactive and dissociates slowly to its active monomeric form after subcutaneous
injection; the lag time for dissociation delays the onset of action
10
. Mutation of ProB28
yields rapid-acting insulins (RAIs) by disrupting contacts that are critical for dimer
formation
8
, but replacement of Pro through conventional mutagenesis also increases the
flexibility and perturbs the trajectory of the protein backbone (Fig. 1B). We sought a means
to disrupt the dimer interface without releasing the conformational constraints characteristic
of proline by using non-canonical amino acid (ncAA) mutagenesis
11
–
13
. Specifically, we
introduced hydroxyl groups at the 4-position of ProB28 (Fig. 1B, C) by replacing Pro with
Hzp or Hyp. In addition to introducing a polar functional group and the capacity for
hydrogen-bonding (including transannular hydrogen bonding), hydroxylation at the 4-
position is known to alter the
endo
/
exo
preference of the pyrrolidine ring and the
cis
/
trans
equilibrium of the backbone amide bond
14
–
16
.
We expressed modified proinsulins (PIs) in the proline-auxotrophic
E. coli
strain CAG18515
in M9 minimal media supplemented with Hyp or Hzp. The extent of replacement of Pro by
either Hyp or Hzp was approximately 90%
17
–
18
as determined by matrix-assisted laser
desorption/ionization mass spectrometry (MALDI-MS; fig. S1). Denatured PIs were purified
by Ni-NTA affinity chromatography in yields of 32 mg/L for Hzp-PI and 29 mg/L for Hyp-
PI (table S1) from the inclusion body fraction. The PIs were refolded and cleaved with
trypsin and carboxypeptidase B
19
. The resulting mature insulins were purified by reversed
phase HPLC
18
–
19
, and proper proteolytic processing of each variant was verified by
MALDI-MS (table S1). Wild-type insulin (ProI) and RAI Aspart (AspI, in which ProB28 is
replaced by aspartic acid) were produced similarly. All of the variants caused similar
reductions in blood glucose upon subcutaneous injection into diabetic mice (Fig. 1D)
2
,
20
–
22
.
RAIs cannot be distinguished from ProI in rodent models
23
.
In the absence of Zn
2+
and phenolic preservatives, insulins dimerize with a dissociation
constant (K
D
) of approximately 10 μM. In contrast, K
D
for RAIs is typically >500 μM, and
it is believed that destabilization of the dimer interface causes the accelerated onset of
insulin action after subcutaneous injection
20
,
24
–
25
. Monomeric forms of insulin give rise to
characteristic circular dichroism (CD) spectra with distinct minima at 208 and 222 nm
26
–
27
(e.g., AspI; Fig. 2A). Dimerization causes a loss of negative ellipticity at 208 nm
27
(e.g.,
ProI; Fig. 2A). At a concentration of 60 μM, Hypl appears to be monomeric (with a CD
spectrum nearly identical to that of AspI; Fig. 2A) while the spectrum of HzpI suggests a
dimeric insulin (Fig. 2A). Sedimentation velocity (SV) and sedimentation equilibrium (SE)
experiments were consistent with the results of the CD analysis (fig. S3). SE data were fitted
to a model of monomer-dimer-hexamer self-association (SEDPHAT)
28
–
29
, and yielded
monomer-dimer dissociation constants (K
D
) of >200 μM and 25 μM for HypI and HzpI,
respectively.
Previous studies of RAIs have shown that destabilization of the dimer interface correlates
with accelerated dissociation of the hexamer and rapid onset of insulin action
8
,
13
. Triggered
dissociation of Zn
2+
-hexamers by addition of terpyridine
30
revealed nearly identical rates of
Lieblich et al.
Page 2
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
dissociation for HypI and ProI, (
τ
1/2
= 87.0 ± 10 s and 90.4 ± 4.2 s, respectively; Fig. 2B and
fig. S4) while HzpI exhibited kinetics similar to those of AspI (
τ
1/2
= 53.6 ± 3.7 s and 42.7
± 4.3 s, respectively; Fig. 2B and fig. S4). We found these results surprising – replacement
of Pro by Hyp destabilizes the dimer but has essentially no effect on hexamer dissociation,
while introduction of Hzp causes little change in dimer stability but a substantial increase in
the rate of hexamer disassembly.
Each of the insulin variants was subjected to fibrillation lag time analysis (Fig. 2C)
31
. We
found similar times to onset of fibrillation for HypI, ProI and AspI; in contrast, HzpI is
markedly more resistant to aggregation, with a mean time to onset more than three-fold
longer than that observed for ProI. The behavior of HzpI is especially striking, in that it
combines fast hexamer dissociation with enhanced stability toward fibrillation.
Each subunit in the insulin hexamer adopts one of two conformational states (T or R),
depending on the concentration of phenolic ligand (Fig. 1A)
13
. Pharmaceutical formulations
are prepared in the more stable R
6
form, whereas the T-state is observed in the absence of
phenolic ligands, most commonly in the form of T
2
-dimers
32
. To elucidate the molecular
origins of the dissociation and fibrillation behavior of HypI and HzpI, we examined crystal
structures of both states.
Hydroxylation at ProB28 does not cause substantial perturbation of the overall insulin
structure (Fig. 3, fig. S5). In comparison with ProI, the backbone RMSD values of HypI and
HzpI are 0.31 Å (T
2
- HypI), 0.44 Å (T
2
- HypI), 0.38 Å (R
6
- HypI) and 0.69 Å (R
6
- HypI)
33
.
The most notable feature of the HzpI structures is the proximity of the hydroxyl group of
Hzp to the backbone carbonyl oxygen atom of GluB21
′
, which lies across the dimer
interface (denoted by prime; Fig. 3B, E). The inter-oxygen distances (2.8 Å in the T
2
structure, 2.7 Å in R
6
), are consistent with the formation of strong hydrogen bonds between
the hydroxyl group of HzpB28 and the backbone carbonyl of GluB21
′
in both structures.
An analogous hydrogen bond has been observed in a structure (PDB ID: 1ZEH) of R
6
-AspI
co-crystallized with
m-
cresol
34
; here the phenolic ligand serves as the hydrogen-bond donor
(fig. S6). Although the significance of this hydrogen bond has not been discussed in the
literature, we suggest that it may play an important role in determining the relative stabilities
of the insulin species involved in dissociation and fibrillation. In contrast to the (4
S
)-
hydroxyl group of Hzp, the (4
R
)-hydroxyl of HypB28 does not contact any
crystallographically resolved hydrogen bond acceptor in the T
2
-structure (Fig. 3C), and
appears to bond to an ordered water molecule in the R
6
-hexamer (Fig. 3F). The absence of
new hydrogen-bonding interactions is consistent with the unaltered dissociation and
fibrillation kinetics of HypI.
Taken together, our results show that replacement of Pro by Hzp at position 28 of the insulin
B-chain introduces a new hydrogen bond across the inter-subunit interface, accelerates
hexamer dissociation and delays the onset of fibrillation (Table 1). We suggest that the
hydrogen bond between Hzp and Glu21
′
may stabilize the dimer relative to the hexamer, or
perhaps reduce the energy of the transition state for the conformational change from the R-
state to T-state, and thereby speed disassembly. The delayed onset of fibrillation may reflect
changes in the structure and dynamics of the HzpI monomer or in the kinetics of fibril
Lieblich et al.
Page 3
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
nucleation. Subtle conformational effects caused by
4S
-substitution on the pyrrolidine ring
may also contribute to the observed behavior.
35
Whether or not these hypotheses are correct,
the results described here demonstrate the power of ncAA mutagenesis to control
functionally relevant biophysical properties of therapeutic proteins. We anticipate that this
approach will find increasing application in the design of antibody-drug conjugates,
bispecific antibodies, and other novel protein therapeutics.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
We thank J. T. Kaiser, P. Nikolovski, S. Russi, S. Virgil, M. Shahgholi, A. Lakshmanan, and the scientific staff of
Beamline 12-2 at the Stanford Synchrotron Radiation Laboratory for assistance. We thank W. Glenn, A. Madhavi,
and T. Hoeg-Jensen for discussions.
Funding Sources
The work was supported by the Novo Nordisk Foundation. Fellowships from Amgen and from the Natural Sciences
and Engineering Research Council of Canada (NSERC, PGS-D) provided partial support for S. A. L. and K. Y. F.,
respectively. J. K. B. C. acknowledges support of the Resnick Sustainability Institute (Caltech). Support by a grant
to H.T.K. from the National Institutes of Health (R01DK099734) is also acknowledged.
Abbreviations
Hzp
(
4S
)-hydroxyproline
Hyp
(
4R
)-hydroxyproline
RAI
Rapid Acting Insulin
ncAA
non-canonical amino acid
PI
proinsulin
HypI
Hyp-Insulin
HzpI
Hzp-Insulin
ProI
Wild-type human insulin
AspI
Insulin Aspart
CD
circular dichroism
SV
sedimentation velocity
SE
sedimentation equilibrium
References
1. Zaykov AN, Mayer JP, DiMarchi RD. Nat Rev Drug Discov. 2016; 15:425–439. [PubMed:
26988411]
Lieblich et al.
Page 4
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
2. Ciszak E, Beals JM, Frank BH, Baker JC, Carter ND, Smith GD. Structure. 1995; 3:615–22.
[PubMed: 8590022]
3. Menting JG, Whittaker J, Margetts MB, Whittaker LJ, Kong GKW, Smith BJ, Watson CJ, Žáková L,
Kletvíková E, Jiráček J, Chan SJ, Steiner DF, Dodson GG, Brzozowski AM, Weiss MA, Ward CW,
Lawrence MC. Nature. 2013; 493:241–5. [PubMed: 23302862]
4. Huang K, Maiti NC, Phillips NB, Carey PR, Weiss MA. Biochemistry. 2006; 45:10278–10293.
[PubMed: 16922503]
5. Menting JG, Yang Y, Chan SJ, Phillips NB, Smith BJ, Whittaker J, Wickramasinghe NP, Whittaker
LJ, Pandyarajan V, Wan Z-l, Yadav SP, Carroll JM, Strokes N, Roberts CT, Ismail-Beigi F, Milewski
W, Steiner DF, Chauhan VS, Ward CW, Weiss MA, Lawrence MC. Proc Natl Acad Sci U S A.
2014; 111:E3395–404. [PubMed: 25092300]
6. Ivanova MI, Sievers SA, Sawaya MR, Wall JS, Eisenberg D. Proc Natl Acad Sci U S A. 2009;
106:18990–18995. [PubMed: 19864624]
7. Freeman JS. J Am Osteopath Assoc. 2009; 109:26–36. [PubMed: 19193822]
8. Birnbaum DT, Kilcomons MA, DeFelippis MR, Beals JM. Pharm Res. 1997; 14:25–36. [PubMed:
9034217]
9. Carpenter MC, Wilcox DE. Biochemistry. 2014; 53:1296–301. [PubMed: 24506168]
10. Bakaysa DL, Radziuk J, Havel HA, Brader ML, Li S, Dodd SW, Beals JM, Pekar AH, Brems DN.
Protein Sci. 1996; 5:2521–2531. [PubMed: 8976561]
11. Pandey AK, Naduthambi D, Thomas KM, Zondlo NJ. J Am Chem Soc. 2013; 135:4333–4363.
[PubMed: 23402492]
12. Liu CC, Schultz PG. Annu Rev Biochem. 2010; 79:413–44. [PubMed: 20307192]
13. Pandyarajan V, Weiss MA. Curr Diab Rep. 2012; 12:697–704. [PubMed: 22983891]
14. Shoulders MD, Kotch FW, Choudhary A, Guzei IA, Raines RT. J Am Chem Soc. 2010;
132:10857–10865. [PubMed: 20681719]
15. Kuemin M, Nagel YA, Schweizer S, Monnard FW, Ochsenfeld C, Wennemers H. Angew Chem,
Int Ed. 2010; 49:6324–6327.
16. Erdmann RS, Wennemers H. Angew Chem, Int Ed. 2011; 50:6835–6838.
17. Kim W, George A, Evans M, Conticello VP. Chembiochem. 2004; 5:928–936. [PubMed:
15239049]
18. Kemmler W, Peterson JD, Steiner DF. J Biol Chem. 1971; 246:6786–6791. [PubMed: 4942325]
19. Min C-K, Son Y-J, Kim C-K, Park S-J, Lee JW. J Biotechnol. 2011; 151:350–356. [PubMed:
21219941]
20. Brems DN, Alter LA, Beckage MJ, Chance RE, DiMarchi RD, Green LK, Long HB, Pekar AH,
Shields JE, Frank BH. Protein Eng. 1992; 5:527–533. [PubMed: 1438163]
21. Sakata N, Yoshimatsu G, Tsuchiya H, Egawa S, Unno M. Exp Diabetes Res. 2012; 2012
22. Pandyarajan V, Phillips NB, Cox GP, Yang Y, Whittaker J, Ismail-Beigi F, Weiss MA. J Biol Chem.
2014; 289:23367–23381. [PubMed: 24993826]
23. Plum A, Agersø H, Andersen L. Drug Metab Dispos. 2000; 28:155–160. [PubMed: 10640512]
24. Attri AK, Fernández C, Minton AP. Biophys Chem. 2010; 148:28–33. [PubMed: 20202737]
25. Antolikova E, Zakova L, Turkenburg JP, Watson CJ, Hanclova I, Sanda M, Cooper A, Kraus T,
Brzozowski AM, Jiracek J. J Biol Chem. 2011; 286:36968–77. [PubMed: 21880708]
26. Nakagawa SH, Tager HS, Steiner DF. Biochemistry. 2000; 39:15826–15835. [PubMed: 11123908]
27. Pocker Y, Biswas SB. Biochemistry. 1980; 19:5043–5049. [PubMed: 7006683]
28. Zhao H, Schuck P. Acta Crystallogr, Sect D: Biol Crystallogr. 2015; 71:3–14. [PubMed: 25615855]
29. Brown PH, Schuck P. Biophys J. 2006; 90:4651–61. [PubMed: 16565040]
30. Rahuel-Clermont S, French CA, Kaarsholm NC, Dunn MF. Biochemistry. 1997; 36:5837–5845.
[PubMed: 9153424]
31. Vinther TN, Norrman M, Strauss HM, Huus K, Schlein M, Pedersen TÅ, Kjeldsen T, Jensen KJ,
Hubálek F. PLoS ONE. 2012; 7:e30882. [PubMed: 22363506]
32. Palmieri LC, Favero-Retto MP, Lourenco D, Lima LM. Biophys Chem. 2013; 173–174:1–7.
Lieblich et al.
Page 5
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
33. Marshall H, Venkat M, Seng NS, Cahn J, Juers DH. Acta Crystallogr, Sect D: Biol Crystallogr.
2012; 68:69–81. [PubMed: 22194335]
34. Smith GD, Ciszak E, Magrum LA, Pangborn WA, Blessing RH. Acta Crystallogr, Sect D: Biol
Crystallogr. 2000; 56:1541–1548. [PubMed: 11092919]
35. DeRider ML, Wilkens SJ, Waddell MJ, Bretscher LE, Weinhold F, Raines RT, Markley JL. J Am
Chem Soc. 2002; 124:2497–2505. [PubMed: 11890798]
Lieblich et al.
Page 6
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Figure. 1. Hydroxyinsulins retain activity
(
A
) Schematic of hexamer disassembly (adapted from mechanism previously described
8
).
Phenolic ligand (Ph), zinc ion (Zn
2+
), insulin monomer (triangle). Darker shading indicates
the R-state of the hexamer. (
B
) Structures of the B-chain
C
-termini of wild-type insulin
(ProI) and RAI Aspart (AspI). (
C
) Chemical structures of L-proline (
1
), (
2S,4S
)-
hydroxyproline (Hzp;
2
), (
2S,4R
)-hydroxyproline (Hyp;
3
). (
D
) Reduction of blood glucose
following subcutaneous injection of 35 μg/kg insulins into streptozotocin-induced diabetic
mice. Glucose levels were measured post-injection via tail vein sampling. ProI, AspI, HzpI,
HypI or vector were formulated as described
22
. Error bars denote one standard deviation (n
= 3).
Lieblich et al.
Page 7
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Figure. 2. Hydroxylation at ProB28 modulates insulin dimerization, dissociation kinetics, and
stability
(
A
) Far UV CD spectra collected on 60 μM insulins in 10 mM phosphate buffer, pH 8.0 at
25°C. (
B
) Insulin hexamer dissociation following sequestration of Zn
2+
by terpyridine.
Zn
2+
-(terpy) signal was monitored at 334 nm and fitted to a mono-exponential decay. HzpI
and HypI contain 10% ProI. The curves for HypI and ProI are indistinguishable in this plot.
(
C
) Representative fibrillation curves for 60 μM insulins (37°C, 960 RPM; n=4). Insulin
fibrils were detected by the rise in Thioflavin T (ThT) fluorescence that accompanies
binding to fibrillar aggregates.
Lieblich et al.
Page 8
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Figure. 3. Crystal structures of HzpI and HypI
In tan (left), wt insulins from PDB (3T2A, 1ZNJ) highlighting the distance between the
carbon atom at the 4
th
position of ProB28 and its closest neighbors, backbone carbonyl
oxygen atoms of GlyB20’ and GluB21 in the T
2
dimer (
A
) and R
6
hexamer (
D
) forms. In
grey (middle), HzpI in the T
2
dimer (
B
) and R
6
hexamer (
E
) forms. In blue (right), HypI in
the T
2
dimer (
C
) and R
6
hexamer (
F
) forms.
Lieblich et al.
Page 9
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Lieblich et al.
Page 10
Table 1
Biophysical Characteristics of Insulin Variants
Protein
Hexamer t
½
(s)
Fibrillation lag time(h)
K
D
dimer (μM)
Prol
90.4 ± 4.2
5.1 ± 1.5
9 μM
Hzpl
*
53.6 ± 3.7
19.6 ± 2.6
~25 μM
Hypl
*
87.0 ± 10
5.5 ± 1.2
>200 μM
Aspl
42.7 ± 4.3
5.3 ± 1.0
>500 μM
Errors are given as one standard deviation (n > 4).
*
10% Prol present.
J Am Chem Soc
. Author manuscript; available in PMC 2018 February 14.