Synthesis of 18F-arenes from spirocyclic iodonium(III) ylides via continuous-flow microfluidics

Article abstract of DOI:10.1016/j.jfluchem.2015.08.006

Spirocyclic hypervalent iodine(III) ylides have proven to be synthetically versatile precursors for efficient radiolabelling of a diverse range of non-activated (hetero)arenes, highly functionalised small molecules, building blocks and radiopharmaceutical

Full text of DOI:10.1016/j.jfluchem.2015.08.006

Journal of Fluorine Chemistry 178 (2015) 249–253
Contents lists available at ScienceDirect
Journal of Fluorine Chemistry
journal homepage: www.elsevier.com/locate/fluor
Short Communication
18
Synthesis of F-arenes from spirocyclic iodonium(III) ylides via
continuous-flow microfluidics
a
Samuel Calderwood a,b, Thomas Lee Collier b,c,d, Veronique Gouverneur ,
´
b,c,
Steven H. Liang b,c, , Neil Vasdev
*
*
a University of Oxford, Chemistry Research Laboratory, 12 Mansfield Road, Oxford, UK
b Division of Nuclear Medicine and Molecular Imaging, Massachusetts General Hospital, 55 Fruit Street, Boston, USA
c Department of Radiology, Harvard Medical School, 55 Fruit Street, Boston, USA
d Advion BioSystems, 10 Brown Road, Suite 101, Ithaca, NY, USA
A R T I C L E I N F O
A B S T R A C T
Article history:
Spirocyclic hypervalent iodine(III) ylides have proven to be synthetically versatile precursors for efficient
radiolabelling of a diverse range of non-activated (hetero)arenes, highly functionalised small molecules,
building blocks and radiopharmaceuticals from [18F]fluoride ion. Herein, we report the implementation
of these reactions onto a continuous-flow microfluidic platform, thereby offering an alterative and
automated synthetic procedure of a radiopharmaceutical, 3-[18F]fluoro-5-[(pyridin-3-yl)ethynyl]ben-
zonitrile ([18F]FPEB) and a routinely used building block for click-radiochemistry, 4-[18F]fluorobenzyl
azide. This new protocol was applied to the synthesis of [18F]FPEB (radiochemical conversion
(RCC) = 68 ꢀ 5%) and 4-[18F]fluorobenzyl azide (RCC = 68 ꢀ 5%; isolated radiochemical yield = 24 ꢀ 0%). We
anticipate that the high throughput microfluidic platform will accelerate the discovery and applications of
18F-labelled building blocks and labelled compounds prepared by iodonium ylide precursors as well as the
production of radiotracers for preclinical imaging studies.
Received 8 July 2015
Received in revised form 4 August 2015
Accepted 6 August 2015
Available online 12 August 2015
Keywords:
Fluorine-18
Iodonium(III) ylide
[
18F]FPEB
4-[18F]Fluorobenzyl azide
PET
18F
ß 2015 Elsevier B.V. All rights reserved.
1. Introduction
change to their parent structures, as they require high-affinity
chelators for complexation.
Positron Emission Tomography (PET) is an established
molecular imaging technique that has applications in clinical
diagnosis and drug development, especially in combination with
anatomical imaging techniques including computed tomography
and magnetic resonance imaging. Fluorine-18 (18F) [1] is widely
regarded as the foremost radionuclide for PET due to the
significant use of 19F in drug design [2], as well as a favourable
decay profile (97% b+ decay to 18O). Furthermore, its relatively
Continuous-flow microfluidic technology can offer significant
advantages forthe synthesis of PET radiotracers, including increased
reproducibility, faster reaction kinetics and rapid reaction optimi-
sation [4]. There have been over 50 reported syntheses of labelled
compounds for PET that exploit the advantages of continuous-flow
microfluidics [5]. Recently, continuous-flow microfluidics was used
for the preparation of [18F]FPEB [6], [18F]T807 [7], and [18F]FMISO [8]
for human use, whilst [18F]Fallypride [9] has comparably been
prepared using a micro-reactor.
long half-life (t = 109.8 min) enables imaging timeframes and
½
multi-centre trials [3], which are not possible with other common
We recently reported a novel procedure for the 18F-fluorina-
tion of hypervalent iodonium(III) ylides to give 18F-arenes from
18F-fluoride [10], that provided efficient regiospecific labelling for
a diverse scope of non-activated functionalised (hetero)arenes.
These include electron-donating motifs, which have typically
been difficult to label with preceding hypervalent iodine(III)
mechanisms, as well as highly functionalised molecules and PET
radiopharmaceuticals. This approach has recently been extended
to the automated production of 3-[18F]fluoro-5-[(pyridin-3-
yl)ethynyl]benzonitrile ([18F]FPEB) in a good radiochemical yield
(15–25%, formulated and ready for injection, n = 3) and validated
for human use, utilising a commercial radiosynthesis module [11].
radionuclides, such as carbon-11 (11C [t = 20.3 min]). Radio-
½
metals that are used for PET imaging purposes can have longer
half-lives (e.g., 64Cu [t = 12.7 h] and 89Zr [t = 78.4 h]), however
½
½
these are generally not suitable for small molecule drugs without
*
Corresponding author at: Division of Nuclear Medicine and Molecular Imaging,
Massachusetts General Hospital, 55 Fruit Street, Boston, USA.
E-mail addresses: liang.steven@mgh.harvard.edu (S.H. Liang),
vasdev.neil@mgh.harvard.edu (N. Vasdev).
http://dx.doi.org/10.1016/j.jfluchem.2015.08.006
0022-1139/ß 2015 Elsevier B.V. All rights reserved.

250
S. Calderwood et al. / Journal of Fluorine Chemistry 178 (2015) 249–253
The goal of the work presented herein was to demonstrate a
4 m (32 m
L) reactor, into which the [18F]TEAF/TEAB (Pump 3 [P3])
translation of our spirocyclic iodonium(III) ylide methodology onto
a continuous-flow microfluidic platform and utilise the high
throughput advantages held by this technology over both manual
and other automated protocols. This work aims to enable rapid
radiochemistry optimisation and synthesis of labelled compounds
for preclinical research.
and substrate solutions (Pump 1 [P1]) were introduced in a 1:1
(P1:P3) ratio. This immediately showedthat anautomated approach
could match the RCCs from a manual approach and consequently a
more exhaustive optimisation of the reaction conditions was rapidly
performed. A comparison of DMF, DMSO and CH3CN (Fig. 1A)
showed DMF to be the preferred solvent, though intriguingly CH3CN
performed better at T ꢁ 200 8C, at which a drop in conversion was
observed in all other solvents. In comparison, CH3CN was dismissed
early in the manual methodology as it had led to relatively low
conversion with this substrate and reaction condition (9 ꢀ 8%, n = 3)
[10] when compared with both DMF and DMSO. We attributed the
lower yield to the lower boiling point of CH3CN as the limiting factor.
A study of five different flow rates through the reactor (20–
As proof of concept, our goals were: (1) to confirm that the
iodonium ylide precursors can be suitably translated to
a
commercial continuous flow reactor system with model
a
substrate; (2) to synthesise and isolate 18F-fluorobenzyl azide, a
building block for high throughput click chemistry and (3) to
synthesise a radiopharmaceutical, with [18F]FPEB as a model
compound, which was previously a challenge to synthesise for
preclinical or clinical work.
100
mL/min) showed that there was no significant change in the
conversion, with a slight increase at slower rates being annulled by
the increased time for the reaction and respective decay of activity
(Fig. 1B). Both the precursor loading and base ([18F]TEAF/TEAB)
loading were optimised consecutively using DMF as the solvent. It
was found that increasing the precursor loading to 16 mg/mL
improved the incorporation of fluoride into 2 across the tempera-
ture range and the drop in conversion, which had been seen when
T > 190 8C, was no longer observed (Fig. 1C).
2. Results and discussion
2.1. Synthesis of model 18F-labelled substrates via iodonium ylide
precursors and continuous flow microfluidics
A commercial continuous-flow microfluidic platform (Nano-
Tek1; Advion, Inc.) [12] was utilised for this study. A model biphenyl
iodonium ylide precursor was explored for initial proof-of-concept
and reaction optimisation using tetraethylammonium [18F]fluoride
([18F]TEAF) as the 18F-fluoride source, formed by the elution of 18F-
fluoride from an anion exchange cartridge with a solution of
tetraethylammonium bicarbonate (TEAB). The precursor, 6,10-
dioxaspiro[4.5]decane-7,9-dion-[1,10-biphenyl-4-iodonium] ylide
(1) was reacted with [18F]TEAF to form 4-[18F]fluorobiphenyl (2),
and was chosen for this study to allow for a direct comparison of
incorporation of 18F-fluoride based on our previous manual
synthesis [10]. The initial conditions, translated from our previous
approach (8 mg/mL 1, 12 mg/mL [18F]TEAF/TEAB, DMF, RCC = 85%)
[10], were subjected to a temperature range (130–220 8C) using a
Decreasing the loading of
[
18F]TEAF/TEAB in the system
increased the RCC at lower temperature, although a reduction in
conversion was observed once the temperature exceeded 160 8C.
Increasing the base loading decreased the conversions across the
temperature range 140–220 8C (Fig. 1D). The effect of the base
loading with a series of other ylides has been observed by us and a
mechanistic study will be published elsewhere.
Final optimised conditions for the radiofluorination of 1 (16 mg/
mL 1, 12 mg/mL [18F]TEAF/TEAB, DMF, 60
m
L/min, P1:P3 1:1, 4 m
(32
m
L) reactor) gave excellent incorporation of 18F-fluoride
(95 ꢀ 1%, T = 200 8C, n = 4) (Scheme 1), and offers an improvement
on the RCC established by the manual procedure (85%) [10].
A
Solvent
B
Rate
100
80
60
40
20
0
100
80
60
40
20
0
DMF
DMSO
CH3CN
20
40
60
80
100
Total Rate (µl/min)
Temperature (oC)
C
Precursor Loading
D
Base Loading
100
80
60
40
20
0
100
80
60
40
20
0
16 mg/mL
8 mg/mL
4 mg/mL
18 mg/mL
12 mg/mL
6 mg/mL
Temperature (oC)
Temperature (oC)
Fig. 1. Optimisation plots for radiolabelling of 4-[18F]fluorobiphenyl (2). Conditions (n = 2, 4 m reactor): (A) 8 mg/mL 1, 12 mg/mL [18F]TEAF/TEAB, 60
12 mg/mL [18F]TEAF/TEAB, DMF, 190 8C; (C) 12 mg/mL [18F]TEAF/TEAB, DMF, 60
L/min; (D) 16 mg/mL 1, DMF, 60 L/min. Incorporation yield, as radiochemical conversion
(RCC), and product identity were determined by radioTLC and radioHPLC, respectively.
mL/min; (B) 8 mg/mL 1,
m
m

S. Calderwood et al. / Journal of Fluorine Chemistry 178 (2015) 249–253
251
Scheme 1. Radiosynthesis of 4-[18F]fluorobiphenyl (2).
As the radiochemical conversion was almost quantitative, we
subjected 6,10-dioxaspiro[4.5]decane-7,9-dion-[1,3,5-trimethyl-
benzene-2-iodonium] ylide (3) to these conditions (16 mg/mL 3,
This allowed for an increased reaction residence time (32–64 s), while
reducing the increase in time to end of synthesis.
The adoption of the 8 metre reactor increased the incorporation
across the temperature range investigated (Fig. 2A), with the
optimum temperature being found to be 210 8C (68 ꢀ 5%, n = 3),
which represents an increase in conversion of 17% under a shorter
residence time than previously published microfluidic techniques
[15].
12 mg/mL [18F]TEAF/TEAB, DMF, 60
mL/min) (Scheme 2), in an
attempt to quantify the degree to which the application on
microfluidics could affect the conversion of these reactions. This
substrate gave a conversion of 82 ꢀ 4% (T = 220 8C, n = 4) to 2-
[
18F]fluoro-1,3,5-trimethylbenzene (4), which is almost double that
obtained by our previously published method (45 ꢀ 13%, n = 3) [10].
These conditions were successfully taken forward and combined
with SPE isolation to give a non-decay corrected RCY = 24 ꢀ 0.4%
(n = 3, relative to the amount of [18F]fluoride obtained at the end of
bombardment) in a totalsynthesis timeof 60 min from target unload to
assay of final product vial, with a radiochemical purity (RCP) of 95%
(Scheme 3). This procedure was fully automated from the azeotropic
drying of the 18F-fluoride to the final collection vial.
The use of these model substrates has shown that the
application of
a microfluidic reactor has the capability for
significantly improving the conversion of spirocyclic iodonium
ylides to 18F-arenes and justified our next efforts to apply this
technology to clinically relevant labelled compounds.
2.2. Synthesis of 4-[18F]fluorobenzyl azide
2.3. Synthesis of 3-[18F]Fluoro-5-[(pyridin-3-yl)ethynyl]benzonitrile
([18F]FPEB)
The first target molecule chosen for evaluation of this
microfluidic protocol was 4-[18F]fluorobenzyl azide (6), which is
used in a ‘building-block’ approach to radiopharmaceuticals which
cannot be directly fluorinated. This is achieved via fast, metal-free
‘‘click’’ reactions with strained alkynes, such as for Bombesin [13],
a GRP-receptor-specific 14 amino acid neuropeptide. Previous
procedures for the synthesis of 6 have followed several methodol-
ogies; via a stepwise procedure (4-steps, 75 min, RCY = 34% [decay
corrected]) [14], nucleophilic aromatic substitution of a nitro
group [13] or via microfluidic approach using diaryliodonium salt
precursors which gave a RCC of 51% with 3% of the labelled anisole
by-product [15]. The stepwise procedure was automated using
solid supports to improve the RCY to up to 60% over a 40 min
synthesis time [16]. We have previously reported a manual one-
step radiosynthesis using an iodonium ylide precursor (6) with
RCY of 25 ꢀ 10% (n = 3) [10] via solid phase extraction (SPE) isolation
techniques, utilising this spirocyclic iodonium(III) ylide chemistry
Initial reactions, using the optimised conditions for the reaction
of 1, on 6,10-dioxaspiro[4.5]decane-7,9-dion-[1-(azidomethyl)-
benzene-4-iodonium] ylide (5) gave moderate conversion to 6
(31 ꢀ 1%, T = 200 8C, n = 3), and a temperature dependence similar to
that seen for radiofluorination of 1 (Fig. 2A). However, unlike in the
synthesis of 2, it was found that there was a significant increase in the
yield when the flow rate was dropped from 60 mL/min to 20 mL/min
(44%, T = 200 8C, n = 1) (Fig. 2B). This is likely attributed to the
increase in residence time in the reactor (32–96 s), though this
decrease in total flow rate consequently increases the time taken per
reaction. In an attempt to enable an increase in residence time, whilst
reducing the obligatory increase in run time, the reactor was changed
from 4 metres in length (32 mL volume) to 8 metres (64 mL volume).
Our final efforts focused on the synthesis of a radiopharmaceuti-
cal targeting the metabotropic glutamate receptor subtype type 5
(mGluR5), a glutamatergicneuralreceptorimplicated ina numberof
central nervous system disorders including chronic neurodegener-
ative disorders, such as Huntington’s and Parkinson’s disease [17].
3-[18F]Fluoro-5-[(pyridin-3-yl)ethynyl]benzonitrile ([18F]FPEB, 8)
is a mGluR5 antagonist that has been successfully applied in
preclinical [18,19] and clinical PET research studies [20,21].
A
100
80
60
40
20
0
8m Reactor
4m Reactor
Temperature (oC)
B
50
40
30
20
10
0
20
40
60
80
100
120
Total Rate (µL/min)
Fig. 2. Temperature variation optimisation for the radiofluorination of 4-
[
18F]fluorobenzyl azide (6). Conditions (n = 1, 4 m reactor; n = 3, 8 m reactor):
(A) 16 mg/mL 5, 12 mg/mL [18F]TEAF/TEAB, DMF, 60
12 mg/mL [18F]TEAF/TEAB, DMF, 200 8C, 4 m reactor.
mL/min; (B) 16 mg/mL 5,
Scheme 2. Radiosynthesis of 2-[18F]fluoro-1,3,5-trimethylbenzene (4).

252
S. Calderwood et al. / Journal of Fluorine Chemistry 178 (2015) 249–253
Scheme 3. One-step radiosynthesis and isolation of 4-[18F]fluorobenzyl azide (6).
Original routes to [18F]FPEB required high temperatures and
decreased to 100 8C. However, no hydrolysis was observed by
radio-HPLC during the optimisation, even at 220 8C, with the sole
radioproduct identified as 8. It was found that there is a relatively
narrow temperature in which an excellent RCC can be achieved
(68 ꢀ 5%, T = 200 8C, n = 2), with relatively rapid decrease in
conversion observed with variation from this point (Scheme 4). To
our knowledge and in our hands, this conversion is superior to all
other known routes to [18F]FPEB. We envision that continuous flow
microfluidics in conjunction with iodonium ylide based precursors
will be widely useful for preparing preclinical doses of [18F]FPEB
from relatively small amounts of 18F-fluoride and should prove
useful for clinical translation.
several radiochemical impurities have been identified during
production, yet only low RCYs were recovered (1–5%) [22,23]. The
presence of an electron-withdrawing nitrile group meta- to the
fluorination centre leads to a displacement of traditional leaving
groups (Cl or NO2) being disfavoured. Our group has published two
routes to [18F]FPEB previously. The first applied the NanoTek1 to
facilitate a SNAr displacement of a nitro group, which gave low
radiochemical yields (2%) [6]. More recently we have extended this
iodonium(III) ylide methodology [11] to give good RCC under a
manual procedure (49 ꢀ 6%, n = 3) and consequent adaption for
automation on a commercial radiofluorination platform led to
[
18F]FPEB formulated for injection in an isolated uncorrected RCY
of 20 ꢀ 5% (n = 3). Hydrolysis of the nitrile group lead to the formation
3. Conclusion
of a radiolabelled by-product and its suppression could increase
radiochemical conversions. It was therefore envisaged that
a
This work has demonstrated the capability and application of
an automated continuous-flow microfluidic platform to the 18F-
fluorination of spirocyclic iodonium(III) ylides, which can improve
incorporation of 18F-fluoride, and reduce reaction time as well as
reagent use. The benefit of using of the iodonium ylide
methodology allows the labelling of arenes displaying electron-
withdrawing groups, including trifluoromethyl, halide, nitro and
ester substituents at the meta (non-activated) positions. However,
a few exceptions for this methodology are expected, such as those
which cannot tolerate the oxidation conditions. These include
precursors with unprotected hydroxy groups, amines and carbox-
ylic acids. The benefits of the use of microfluidics include the rapid
optimisation using minimal reagents, low activity, over a wide
range of temperatures and residence times. Typically, there is an
increase in radiochemical yield using microfluidics due to the
improved mixing and improved heat transfer capabilities. As proof
of concept, two important 18F-labelled compounds, 4-[18F]fluor-
obenzyl azide and [18F]FPEB, were radiolabelled via this technolo-
gy and should have widespread implications in radiotracer
production.
combination of these two approaches (iodonium ylide chemistry in
conjunction with continuous flow microfluidics) would lead to a
higher yielding process with potentially less impurities to produce
this important radiopharmaceutical.
Following the discovery of the benefits of using the 8 m reactor
on the azide system, this was immediately applied to the
radiofluorination of 3-((7,9-dioxo-6,10-dioxaspiro[4,5]decan-8-
ylidene)-l3-iodanyl)-5-(pyridine-2-ylethynyl)benzonitrile (7)
(Fig. 3). Due to hydrolysis of the nitrile being known, especially
at higher temperatures, the lower temperature range limit was
100
80
60
40
20
0
4. Experimental
Temperature (oC)
Substrates (1, 3, 5 and 7) and the reference compounds ([19F]2,
Fig. 3. Temperature variation optimisation for the radiofluorination of 4-[18F]FPEB
(8). Conditions (n = 2, 8 m reactor): 16 mg/mL 7, 12 mg/mL [18F]TEAF/TEAB, DMF,
[
19F]4, [19F]6 and [19F]8) were synthesised as previously published
[10,11]. Full details of radiochemical synthesis of compounds 2, 4,
6 and 8, including NanoTek1 schematic and automation macros,
can be found in the ESI.
60
mL/min.
Acknowledgements
Financial support was provided by the MRC (SC), Advion (SC &
NV) and Ground Fluor Pharmaceuticals (NV). VG holds a Royal
Society Wolfson Research Merit Award (2013–2018). We thank Dr.
Benjamin H. Rotstein and Dr. Nickeisha Stephenson for selected
substrate synthesis and helpful discussions. We thank Dr Jack A.
Correia, David F. Lee, Jr. and Tim Beaudoin and the Massachusetts
General Hospital PET Core facility for 18F-fluoride production.
Scheme 4. Radiosynthesis of [18F]FPEB (8).

S. Calderwood et al. / Journal of Fluorine Chemistry 178 (2015) 249–253
253
[11] N.A. Stephenson, J.P. Holland, A. Kassenbrock, D.L. Yokell, E. Livni, S.H. Liang,
Appendix A. Supplementary data
N. Vasdev, J. Nucl. Med. 56 (2015) 489–492.
[12] G. Pascali, L. Matesic, T. Collier, N. Wyatt, B. Fraser, T. Pham, P. Salvadori, I.
Greguric, Nat. Protoc. 9 (2014) 2017–2029.
[13] L. Campbell-Verduyn, L. Mirfeizi, A. Schoonen, R. Dierckx, P. Elsinga, B. Feringa,
Angew. Chem. Int. Ed. 50 (2011) 11117–11120.
[14] D. Thonon, C. Kech, J. Paris, C. Lemaire, A. Luxen, Bioconjug. Chem. 20 (2009)
817–823.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jfluchem.2015.
08.006.
[15] J. Chun, V. Pike, Eur. J. Org. Chem. 24 (2012) 4541–4547.
[16] C. Lemaire, L. Libert, A. Plenevaux, J. Aerts, X. Franci, A. Luxen, J. Fluor. Chem.
138 (2012) 48–55.
References
[1] L.S. Cai, S.Y. Lu, V.W. Pike, Eur. J. Org. Chem. 17 (2008) 2853–2873.
[2] F. Menaa, B. Menaa, O.N. Sharts, J. Mol. Pharm. Org. Process Res. 1 (2013) 104.
[3] P.W. Miller, N.J. Long, R. Vilar, A.D. Gee, Angew. Chem. Int. Ed. 47 (2008)
8998–9033.
[17] F. Bordi, A. Ugolini, Prog. Neurobiol. 59 (1999) 55–79.
[18] J.M. Rook, M.N. Tantawy, M.S. Ansari, A.S. Felts, S.R. Stauffer, K.A. Emmitte,
R.M. Kessler, C.M. Niswender, J.S. Daniels, C.K. Jones, C.W. Lindsley, P.J. Conn,
Neuropsychopharmacology 40 (2015) 755–765.
[4] C. Rensch, A. Jackson, S. Lindner, R. Salvamoser, V. Samper, S. Riese, P.
Bartenstein, C. Wangler, B. Wangler, Molecules 18 (2013) 7930–7956.
[5] G. Pascali, P. Watts, P.A. Salvadori, Nucl. Med. Biol. 40 (2013) 776–787.
[6] S. Liang, D. Yokell, R. Jackson, P. Rice, R. Callahan, K. Johnson, D. Alagille, G.
Tamagnan, T. Collier, N. Vasdev, Med. Chem. Commun. 5 (2014) 432–435.
[7] S. Liang, D. Yokell, M. Normandin, P. Rice, R. Jackson, T. Shoup, T. Brady, G. El
Fakhri, T. Collier, N. Vasdev, Mol. Imaging 13 (2014) 1–5.
[8] M. Zheng, L. Collier, F. Bois, O. Kelada, K. Hammond, J. Ropchan, M. Akula, D.
Carlson, G. Kabalka, Y. Huang, Nucl. Med. Biol. 42 (2015) 578–584.
[9] A. Lebedev, R. Miraghaie, K. Kotta, C. Ball, J. Zhang, M. Buchsbaum, H. Kolb, A.
Elizarov, Lab Chip 13 (2013) 136–145.
[19] J. Wang, W. Tueckmantel, A. Zhu, D. Pellegrino, A. Brownell, Synapse 61 (2007)
951–961.
[20] J. Sullivan, K. Lim, D. Labaree, S. Lin, T. McCarthy, J. Seibyl, G. Tamagnan, Y.
Huang, R. Carson, Y. Ding, E. Morris, J. Cereb. Blood Flow Metab. 33 (2013)
532–541.
[21] D. Wong, R. Waterhouse, H. Kuwabara, J. Kim, J. Brasic, W. Chamroonrat, M.
Stabins, D. Holt, R. Dannals, T. Hamill, P. Mozley, J. Nucl. Med. 54 (2013) 388–396.
[22] T. Hamill, S. Krause, C. Ryan, C. Bonnefous, S. Govek, T. Seiders, N. Cosford, J.
Roppe, T. Kamenecka, S. Patel, R. Gibson, S. Sanabria, K. Riffel, W. Eng, C. King,
X. Yang, M. Green, S. O’Malley, R. Hargreaves, H. Burns, Synapse 56 (2005)
205–216.
[10] B. Rotstein, N. Stephenson, N. Vasdev, S. Liang, Nat. Commun. 5 (2014) 4365.
[23] K. Lim, D. Labaree, S. Li, Y. Huang, Appl. Radiat. Isot. 94 (2014) 349–354.