Intracellular disulfide reduction by phosphine-borane complexes: Mechanism of action for neuroprotection

Article abstract of DOI:10.1016/j.neuint.2016.05.014

Phosphine-borane complexes are novel cell-permeable drugs that protect neurons from axonal injury in vitro and in vivo. These drugs activate the extracellular signal-regulated kinases 1/2 (ERK1/2) cell survival pathway and are therefore neuroprotective, b

Full text of DOI:10.1016/j.neuint.2016.05.014

Neurochemistry International 99 (2016) 24e32
Contents lists available at ScienceDirect
Neurochemistry International
journal homepage: www.elsevier.com/locate/nci
Intracellular disulfide reduction by phosphine-borane complexes:
Mechanism of action for neuroprotection
Nicholas J. Niemuth a, Alex F. Thompson a, Megan E. Crowe a, Christopher J. Lieven a,
, b, *
Leonard A. Levin a
a Department of Ophthalmology and Visual Sciences, University of Wisconsin School of Medicine and Public Health, United States
b Departments of Ophthalmology and Neurology, McGill University, Canada
a r t i c l e i n f o
a b s t r a c t
Article history:
Phosphine-borane complexes are novel cell-permeable drugs that protect neurons from axonal injury
in vitro and in vivo. These drugs activate the extracellular signal-regulated kinases 1/2 (ERK1/2) cell
survival pathway and are therefore neuroprotective, but do not scavenge superoxide. In order to un-
derstand the interaction between superoxide signaling of neuronal death and the action of phosphine-
borane complexes, their biochemical activity in cell-free and in vitro assays was studied by electron
paramagnetic resonance (EPR) spectrometry and using an intracellular dithiol reporter that becomes
fluorescent when its disulfide bond is cleaved. These studies demonstrated that bis(3-propionic acid
methyl ester) phenylphosphine-borane complex (PB1) and (3-propionic acid methyl ester)
diphenylphosphine-borane complex (PB2) are potent intracellular disulfide reducing agents which are
cell permeable. EPR and pharmacological studies demonstrated reducing activity but not scavenging of
superoxide. Given that phosphine-borane complexes reduce cell injury from mitochondrial superoxide
generation but do not scavenge superoxide, this implies a mechanism where an intracellular superoxide
burst induces downstream formation of protein disulfides. The redox-dependent cleavage of the disul-
fides is therefore a novel mechanism of neuroprotection.
Received 15 January 2016
Received in revised form
23 May 2016
Accepted 31 May 2016
Available online 2 June 2016
Keywords:
Phosphine-borane complexes
Disulfide reduction
Neuroprotection
© 2016 Elsevier Ltd. All rights reserved.
1. Introduction
superoxide anion is a signaling molecule for death of a specific
central nervous system (CNS) neuron, the retinal ganglion cell
Reactive oxygen species (ROS) cause cell death via two disparate
mechanisms, direct damage to macromolecules and activation of
intracellular pathways that transduce a cell death signal (Simon
et al., 2000). In previous studies we demonstrated that
(RGC), after injury to its axon contained within the optic nerve
(Kanamori et al., 2010a, 2010b; Lieven et al., 2006; Scott et al.,
2010). Despite near-complete reduction of intracellular levels of
superoxide by scavengers such as pegylated superoxide dismutase
(PEG-SOD) (Kanamori et al., 2010a) and metallocorroles (Catrinescu
et al., 2012; Kanamori et al., 2010b), there is much less neuro-
protection of neurons in animal models of optic neuropathy. This
disparity between the degree of observed signaling by superoxide
and the incomplete neuroprotection when superoxide is scavenged
has three possible explanations.
First, the superoxide burst could be a result of apoptosis and not
its cause, e.g. if cytochrome c release during apoptosis caused
reduction of proximal intermediates in the mitochondrial electron
transport chain. The reduced intermediates would react with mo-
lecular oxygen to produce superoxide (Cai and Jones, 1998). Against
this possibility are data from in vitro (Lieven et al., 2012) and lon-
gitudinal in vivo studies (Kanamori et al., 2010a) demonstrating
that the superoxide burst in fact precedes cytochrome c release and
Abbreviations: CNS, central nervous system; DABCO, 1,4-diazobicyclo[2.2.2]oc-
tane; DEPMPO, 5-(diethoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide; DTNB, 5,5’-
dithiobis-(2-nitrobenzoic acid); DTT, dithiothreitol; EPR, electron paramagnetic
resonance; ERK1/2, extracellular signal-regulated kinases 1/2; P1, bis(3-propionic
acid methyl ester) phenylphosphine; P2, (3-propionic acid methyl ester)diphenyl-
phosphine; PAMPA-BBB, parallel artificial membrane permeation assay for the
prediction of blood-brain barrier penetration; PAMPA, parallel artificial membrane
permeation assay for the prediction of blood-brain barrier penetration; PB1, bis(3-
propionic acid methyl ester) phenylphosphine-borane complex; PB2, (3-propionic
acid methyl ester) diphenylphosphine-borane complex; PEG-SOD, poly(ethylene
glycol)-conjugated superoxide dismutase; RGC, retinal ganglion cell; ROS, reactive
oxygen species; TCEP, tris(2-carboxyethyl)phosphine.
* Corresponding author. Department of Ophthalmology and Visual Sciences,
University of Wisconsin Medical School, 1300 University Ave, Room 694, Madison,
WI 53792, United States.
http://dx.doi.org/10.1016/j.neuint.2016.05.014
0197-0186/© 2016 Elsevier Ltd. All rights reserved.

N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
25
phosphatidylserine externalization, respectively. Second, there are
oxidation-independent stress pathways such as the endoplasmic
reticulum stress signaling pathways, that operate independently of
superoxide signaling. Third, the intracellular superoxide burst
could be rapidly followed by activation of the downstream effects
of superoxide, e.g. oxidation of one or more critical signaling
macromolecules. If scavenging superoxide occurs after the down-
stream pre-apoptotic pathways are activated, then levels of neu-
roprotection are likely to be incomplete.
injury, optic nerve transection and experimental glaucoma, with a
biological mechanism of action that involves activation of the
extracellular signal-regulated kinases 1/2 (ERK1/2) pathway
(Almasieh et al., 2011).
Although phosphine itself is toxic, alkyl phosphines are far less
toxic. The LD50 value of TCEP in rats dosed orally is 3500 mg/kg, and
an LD50 greater than 1024 mg/kg when administered by i.p injec-
tion (Hampton Research, 2016). Previous toxicity testing of PB1 and
PB2 in vitro and PB1 in vivo demonstrated no toxicity to RGCs and
One such downstream target for superoxide induced by axot-
omy is oxidation of cysteine thiols, with consequent formation of
disulfide bonds that modify protein structure and function (Carugo
et al., 2003; Park and Raines, 2001). Previous studies demonstrated
that the disulfide reducing agent dithiothreitol (DTT) can increase
in vitro survival of CNS neurons in mixed retinal culture (Geiger
et al., 2002). Likewise, studies with tris(2-carboxyethyl)phosphine
retinal endothelial cells up to 1 mM with PB1 and 100 mM with PB2,
which are 105 and 108 times the optimal reducing concentration for
neuroprotection (Schlieve et al., 2006).
The phosphine-borane complexes PB1 and PB2 are neuro-
protective in vitro and in vivo, and their structural similarity to TCEP
is consistent with an ability to reduce disulfide bonds. Yet their
biochemical mechanism of action could result from either scav-
enging of the upstream superoxide burst that signals neuronal
death or reduction of intracellular disulfides. Inhibition of either
pathway would be associated with increased neuronal survival. In
order to distinguish these two possibilities, electron paramagnetic
resonance, intracellular disulfide reducing probes, and superoxide
assays were used to assess the biochemical effects of these neu-
roprotective molecules. Their pharmacological characteristics were
tested in order to determine whether they would be able to
penetrate the blood-brain barrier and cell membranes, consistent
with in vivo neuroprotective activity. Their reducing activity was
compared to DTT and TCEP at various pH to determine their activity
at a biologically relevant pH. Phosphine-borane complexes were
potent intracellular reducing agents, with pharmacological and
pharmaceutical properties that would predict activity as CNS
neuroprotectants.
(TCEP),
a disulfide-reducing phosphine, demonstrated that
reversing sulfhydryl oxidation prevents neuronal death after
axotomy in vivo (Geiger et al., 2002) and after optic nerve crush in
rats (Swanson et al., 2005). Such results are consistent with disul-
fide formation being a downstream pathway for cell death induced
by axonal injury.
Based on the observed neuroprotection with DTT and TCEP, we
synthesized membrane permeable derivatives of TCEP, phosphine-
borane
complexes
bis(3-propionic
acid
methyl
ester)
phenylphosphine-borane complex (PB1) and (3-propionic acid
methyl ester) diphenylphosphine-borane complex (PB2) (Fig. 1).
These compounds have a positive, dose-dependent effect on
neuronal viability after axonal injury at concentrations much lower
than that of non-derivatized TCEP (Schlieve et al., 2006). PB1 and
PB2 are neuroprotective in vivo in two rat models of CNS axonal
Fig. 1. Biochemistry of phosphine-borane complexes used in the studies. (A) bis(3-propionic acid methyl ester) phenylphosphine-borane complex (PB1); (B) (3-propionic acid
methyl ester) diphenylphosphine-borane complex (PB2); (C) bis(3-propionic acid methyl ester) phenylphosphine (P1); (D) (3-propionic acid methyl ester) diphenylphosphine (P2).
(E) Mechanism of reduction of disulfide by phosphine in water, modified from Burns et al. (1991). Panels A and B are redrawn from Schlieve et al. (2006).

26
N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
2. Material and methods
micrographs images were longitudinally captured over 60 min with
an Axiovert 135 microscope (Carl Zeiss, Dublin, CA) equipped with a
Nikon D5000 DSLR camera (Nikon, Melville, NY), and analyzed off-
line using NIH ImageJ. Images were processed to yield mean fluo-
rescence of individual cells. Images were processed to identify in-
dividual cells with the ImageJ Auto Threshold plug-in, choosing the
thresholding algorithm selected on a picture-by-picture basis be-
tween Li (22 images), Moments (32 images), Huang (1 image), and
Triangle (1 image) in order to achieve the greatest combination of
cell detection and separation of individual cells. The intensities of
the identified cells were then measured, and non-normalized in-
tensities averaged across three separate experiments to assess the
extent of probe reduction.
2.1. Chemicals
Porcine polar brain lipids were obtained from Avanti Polar Lipids
(Alabaster, AL). Dithiothreitol (DTT), 1,4-diazobicyclo[2.2.2]octane
(DABCO), and dodecane were obtained from Acros Organics (Geel,
Belgium). Poly(ethylene glycol)-conjugated superoxide dismutase
(PEG-SOD), xanthine, xanthine oxidase, 5,5’-dithiobis-(2-
nitrobenzoic acid) (DTNB), tris(2-carboxyethyl)phosphine (TCEP),
and dimethylformamide were from Sigma (St Louis, MO). Hydro-
ethidine was from Invitrogen (Carlsbad, CA). The fluorescent dithiol
probe DSSA-1 was a kind gift of Dr. Daniel Sem of Marquette Uni-
versity (Milwaukee, WI). 5-(diethoxyphosphoryl)-5-methyl-1-
pyrroline-N-oxide (DEPMPO) was from Radical Vision (Marseilles,
France). Unless otherwise stated, all other reagents were from
Fisher Scientific (Pittsburgh, PA).
2.5. Superoxide detection by oxidation of hydroethidine
For fluorescence assays, superoxide was generated by reaction
of xanthine (1 mM) with xanthine oxidase (0.05 U/mL) and
detected by fluorescence of its reaction product with hydroethidine
(1 mM), 2-hydroxyethidium (excitation 396 nm, emission 590 nm),
on a Wallac 1420 VICTOR 2 T Multilabel Counter (PerkinElmer, Inc.,
Wellesley, MA). Reactions were carried out at room temperature in
the presence or absence of PEG-SOD, PB1, or PB2.
2.2. Synthesis of phosphines
PB1 and PB2 were synthesized according to our published
methods (Schlieve et al., 2006) at the Keck-University of Wisconsin
Comprehensive Cancer Center Small Molecule Screening Facility
(Madison, WI). Unprotected phosphines bis(3-propionic acid
methyl ester) phenylphosphine (P1) and (3-propionic acid methyl
ester)diphenylphosphine P2, identical to PB1 and PB2 in all but the
presence of the protective borane group, were synthesized as fol-
lows: P1 was synthesized as the stable intermediate of PB1 syn-
thesis (Schlieve et al., 2006). P2 was synthesized by a method
analogous to the published method for PB2 synthesis. Diphenyl-
phosphine (0.190 g) was dissolved in methanol (8 ml) in a flame-
dried round bottom flask under argon at room temperature. Po-
tassium hydroxide was added to this mixture, followed by the
drop-wise addition of methyl acrylate (0.108 mL). The reaction
mixture was allowed to stir at room temperature for 6 h, after
which the methanol was removed en vacuo. The white residue was
taken up in dichloromethane (10 mL) and washed with 0.5 N HCl
(1 ꢀ 5 mL) and brine (1 ꢀ 5 mL). The aqueous layers were washed
with dichloromethane (10 mL), and the combined organic layers
were dried over MgSO4(s), filtered, and concentrated in vacuo. The
residue was purified by flash chromatography (silica gel, 30% v/v
ethyl acetate in hexanes). P2 was isolated as a pale yellow oil
(0.219 g, 0.76 mmol, 76% yield).
2.6. Superoxide detection by electron paramagnetic resonance
Superoxide was generated by reaction of hypoxanthine
(200 mM) with xanthine oxidase (0.0025 U/mL) in the presence of
the spin trap DEPMPO. A reference spectrum for the hydroxyl
adduct was generated through the Fenton reaction of FeSO4
(100
mM), diethylenetriaminepentacetic acid (100 mM), and
hydrogen peroxide (1 mM) in the presence of DEPMPO (100 mM).
The formation of superoxide-DEPMPO and hydroxyl-DEPMPO ad-
ducts were monitored using electron paramagnetic resonance
(EPR) spectrometry. EPR spectra were recorded at room tempera-
ture using a Bruker EMX spectrometer operating at 9.85 GHz and
equipped with a Bruker ER 4119HS-WI high sensitivity resonator.
Typical spectrometer parameters were: scan range 100 G, time
constant 1.28 ms, sweep time 42 s, modulation amplitude 1.0 G,
modulation frequency 100 kHz, microwave power 20 mW. Micro-
pipettes (50 mL) were used as sample tubes. Spectra were averaged
over 5 scans. Observations were repeated 2e3 times.
2.7. Aqueous solubility
2.3. Measurement of In vitro reducing activity
One mg of pure compound was dissolved in 10 mL ultrapure
water and mixed by inversion for 72 h. Samples were taken at 48
and 72 h, and solubility was determined by detection of PB1 and
PB2 in solution using high performance liquid chromatography
(ThermoFinnigan SpectraSYSTEM P4000 quaternary pump;
AS3000 autosampler, UV2000 detector; and Shimadzu CR501
Solid PB1 and PB2 were dissolved in nitrogen-purged 10 mM
DABCO in dimethylformamide to a final concentration of 10 mM
and borane-deprotected by refluxing at 60 ꢁC for 2 h. Ellman’s re-
agent (5,5’-dithiobis-(2-nitrobenzoic acid), DTNB) at 30 mM in
methanol was reacted with various concentrations of DTT, TCEP,
PB1, or PB2 in buffer at pH 5, 7, and 9 in a 96-well plate for 5 min at
room temperature. Reduction of DTNB was detected by spectro-
photometry at 405 nm using a ThermoMax plate reader (Molecular
Devices, Sunnyvale, CA).
Chromatopac integrator. Phenomenex Luna C18, 5
m, 250 ꢀ 4.6 mm
column) with the following program: 5% acetonitrile/95% ddH2O
0e5 min, 50% acetonitrile/50% ddH2O 15e45 min, 5% acetonitrile/
95% ddH2O 50 min-hold.
2.4. Measurement of intracellular reducing activity
2.8. Octanol/water partitioning
RGC-5 cells (gift of Neeraj Agarwal, Ph.D.) are a cell line derived
from retina neurons that despite their name are most similar to the
661W cone photoreceptor cell line, a type of retinal neuron (Al-
Ubaidi, 2014; Krishnamoorthy et al., 2013; Thompson et al., 2015).
They were incubated in 20 mM DSSA-1 for 1 h. Cells were thor-
oughly rinsed 5 times with Hanks balanced salt solution to remove
The method of Sohn and Park (2003) was used with minor
modification to determine the partitioning coefficients of PB1 and
PB2. Equal volumes of octanol and milliQ deionized water were
mixed by inversion for 30 min and allowed to equilibrate overnight
at room temperature. PB1 or PB2 (1 mg) were dissolved in 5 mL of
equilibrated octanol and mixed by inversion with an equal amount
of equilibrated water overnight. Samples were allowed to
extracellular probe, and PB1, PB2 or TCEP added. Serial fluorescent

N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
27
equilibrate for 24 h, and then concentrations in octanol were
determined by absorbance at 265 nm using a NanoDrop-1000
spectrophotometer. Partitioning coefficients were determined by
comparison to absorbance of unmixed octanol controls.
2.9. Blood-brain barrier permeability
A parallel artificial membrane permeation assay for the pre-
diction of blood-brain barrier penetration (PAMPA-BBB) was used
to determine log Pe for predicted blood-brain barrier permeability
of PB1 and PB2. Several groups have shown accurate prediction of
blood-brain barrier permeability using this assay (Di et al., 2003).
Artificial membranes were created on a 96-well filter plate using
porcine polar brain lipids dissolved in dodecane (20 mg/mL). PB1 or
PB2 (200
mM) dissolved in 200
mL milliQ deionized water was
placed in the donor well above the membrane, and 300
mL pure
water was placed in the acceptor well below the membrane. Con-
centration of compound in acceptor wells after 18 h at room tem-
perature was assessed by absorbance at 265 nm using a NanoDrop-
1000 spectrophotometer. Comparison to absorbance of equilibrium
mixtures of 200 mL PB1 or PB2 solutions with 300 mL water were
then used to determine log Pe. Three independent experiments
were performed, each with 3e6 replicates.
2.10. Accelerated stability
The method of Sohn and Park (2003) was used with minor
modification to determine the accelerated stability of PB1 and PB2.
Pure PB1 or PB2 (4 mg) were incubated in sand baths at 40 ꢁC and
60 ꢁC for 37 and 125 days. 0.5 mg from each sample was dissolved
in acetonitrile and concentration of undegraded compound deter-
mined by high performance liquid chromatography as above.
2.11. Statistical analyses
Comparison between two groups was by two-sample Student’s
t-test. Comparison of a group with a population mean was by one-
sample Student’s t-test. Comparison of more than 2 groups was by
ANOVA followed by post hoc testing with Tukey-Kramer HSD. All
data are reported as mean SEM.
Fig. 2. Phosphine-borane complexes are disulfide reducing agents at physiological
pH. PB1, PB2, DTT, and TCEP (0.1, 0.2, 0.5, 1, 2, 3, 5 mM) were reacted with 7.5 mM DTNB
at pH 5, 7, and 9. While all four compounds reduced disulfides more effectively at pH 5
and pH 9, PB2 reduced disulfides more effectively than DTT or TCEP at the biologically
relevant pH of 7. Results are representative of three independent experiments.
3. Results
3.1. Reduction of disulfide bonds in cell-free assays
Phosphine-borane complexes were hypothesized to reduce di-
sulfide bonds, similar to the activity seen with the parent molecule
TCEP. Such an activity could explain the in vivo efficacy of PB1 and
PB2 in optic nerve transection and experimental glaucoma
(Almasieh et al., 2011). To study this, the protective borane group
was chemically removed with DABCO (Brisset et al., 1993) and the
in vitro reactivity of deprotected PB1 and deprotected PB2
measured spectrophotometrically with DTNB, a colorless disulfide
that becomes yellow when reduced). There was a linear relation
between concentration and reducing power for each compound
(Fig. 2, representative of 3 independent experiments). Deprotected
PB2 had the highest reducing capability at pH 7, possibly explaining
its activity at very low (picomolar to nanomolar concentrations) in
cultured neurons (Schlieve et al., 2006). Deprotected PB1 was a less
potent reducing agent than PB2, but was minimally affected by
changes in pH values. The non-deprotected (borane-complexed)
forms PB1 and PB2 did not reduce disulfides in cell-free assays.
3.2. Intracellular reduction of disulfide bonds in vitro
Given that PB1 and PB2 were chemically able to reduce disul-
fides when deprotected, the ability of these phosphines to enter
cells and reduce disulfides when the borane was present was
studied. To establish whether PB1 and PB2 were able to reduce
disulfide bonds of proteins within cells, the intracellular redox
sensitive probe DSSA-1 (Pullela et al., 2006) was used. DSSA-1 be-
comes fluorescent (excitation 485e495 nm, emission 516e525 nm)
when reduced (Fig. 3A). Cells were loaded with DSSA-1 and incu-
bated with to PB1, PB2, or TCEP. Addition of 10
resulted in an increase in mean cell fluorescence intensity
compared to DSSA alone (26.7 0.3 vs. 22.3 0.2; p < 0.001;
Fig. 3BeC, based on mean of 3 independent experiments). A 1 h
incubation with PB2 (10 M) induced a greater increase in intra-
cellular reduction of DSSA-1 (36.6 0.2 vs. 22.3 0.2; p < 0.001).
mM PB1 for 1 h
m

28
N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
Fig. 3. Phosphine-borane complexes are intracellular disulfide reducing agents in vitro. (A) DSSA-1 is a redox-sensitive probe, which fluoresces upon reduction of its disulfide
linkage. (B) Cells preloaded with DSSA-1 showed a significant increase in fluorescence upon treatment with PB1 and PB2 after 1 h. (C) Representative micrographs of DSSA-1
fluorescence after treatment with PB1, PB2, TCEP, and media alone. Results are the means of three independent experiments SEM. Differences in background reflect reduc-
tion of DSSA-1 in the medium, and not differences in camera exposure.
Treatment with 1 mM TCEP for 1 h resulted in an increase in
intracellular brightness (25.6 0.2 vs. 22.3 0.2; p < 0.001) that
was less pronounced than what was observed with PB1 or PB2,
despite a 100-fold increase in concentration of reductant.
3.4. Pharmacological properties
Despite demonstration of disulfide-reducing activity by these
phosphine-borane complexes in cell-free and in vitro models, there
was little evidence that their pharmacological properties were
consistent with their observed neuroprotective activity in vivo in
models of CNS axonal injury, optic nerve transection and experi-
mental glaucoma (Almasieh et al., 2011). To address this disparity,
their pharmacological properties relevant to CNS activity and use as
clinical neuroprotectants were studied.
The ability to penetrate the blood-brain barrier is critical to
pharmacological activity in the CNS. Using an artificial membrane
system of porcine brain lipids (PAMPA-BBB), log Pe was as
ꢂ4.93 0.04 for PB1 and e5.00 0.06 for PB2. Both values are
significantly (p ¼ 0.008 and 0.025, respectively) more permeable
than the threshold (ꢂ5.34) designated as CNS-penetrant (Di et al.,
2003). Water solubility critically influences the route by which a
drug can be administered as well as its absorption and distribution
in the body. Detection of compound dissolved in water over 72 h
was by high performance liquid chromatography. PB1 was soluble
3.3. Superoxide scavenging
Reactivity with endogenous superoxide anion is a theoretical
alternative mechanism of action for the neuroprotective effect of
phosphine-borane complexes. To test this hypothesis, superoxide
was generated using xanthine/xanthine oxidase and detected by
fluorescence of 2-hydroxyethidium. PB1 and PB2 did not appre-
ciably scavenge superoxide when compared to PEG-SOD control
(Fig. 4A, representative of 3 independent experiments), confirming
previous experiments (Seidler et al., 2010).
Since 2-hydroxyethidium fluorescence is not completely specific
for the reaction product of HEt and superoxide (Zhao et al., 2005),
electron paramagnetic resonance (EPR) was used as an alternate
method for detecting superoxide. P1 and P2, corresponding to PB1
and PB2 but without the protective borane groups, were synthe-
sized and their effect on superoxide measured by EPR. The spin trap
DEPMPO was used to detect superoxide generated by xanthine/
xanthine oxidase. This produces a signal characteristic of the su-
peroxide adduct (DEPMPO/OOH) (Fig. 4B), and is readily distin-
guishable from the hydroxyl adduct generated by the Fenton
reaction (Fig. 4C). The presence of SOD (30 U/mL) eliminated the
signal from superoxide (Fig. 4D). Addition of P1 or P2 did not
diminish the spectra of superoxide (Fig. 4E and F), and instead, the
observed spectra were identical to that of DEPMPO/OH, the hy-
droxyl adduct of DEPMPO. The same result was seen with the
addition of catalase, indicating that the hydroxyl-DEPMPO adduct
was not generated by the Fenton reaction from H2O2. Results are
representative of 2 independent experiments. Together, these re-
sults indicate that these phosphines are reducing agents and not
superoxide dismutase mimetics, with activity analogous to the
glutathione-dependent reduction of a hydroperoxide to an alcohol
(Karoui et al., 1996).
in water up to 237.8
mM, and PB2 was soluble in water up to 3.7 mM,
demonstrating limited water solubility.
Hydrophobicity is another important parameter for structure-
activity relations and biological activity of a compound. Using
octanol/water partitioning, log Poct/wat was 0.6 (somewhat hydro-
phobic) for PB1 and 1.1 (highly hydrophobic) for PB2, signifying
high potential biological activity. Long-term stability is correlated
with accelerated stability at elevated temperatures. Heated samples
were maintained over a period of 125 days and then analyzed by
high performance liquid chromatography. Peak areas for PB1 at
40 ꢁC and 60 ꢁC were 80% and 72% of control respectively at 37 days.
PB2 peak areas were 100% of control for both conditions at 37 days.
At 125 days, the peak area for PB1 at 40 ꢁC was 58% and PB2 was
86%; at 60 ꢁC peak areas for PB1 and PB2 were 12% and 14%
respectively (Fig. 5, representing 1 experiment). Extrapolation us-
ing the Arrhenius equation yielded a calculated half-life for solid
PB1 of 2 years and for solid PB2 of 18 years at 20 ꢁC. Finally, both

N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
29
Fig. 4. PB1 and PB2 do not scavenge superoxide. (A) PB1 and PB2 at 1, 10, and 100 mM, and PEG-SOD at 1, 10, and 100 U/mL were allowed to react with superoxide radical
generated by the reaction of 0.05 U/mL xanthine oxidase (XO) with 1 mM xanthine (X). Superoxide was measured as fluorescence of its reaction product with hydroethidine, 2-
hydroxyethidium, at 465 nm and results are reported as the mean SEM. Asterisk indicates p < 0.05 compared to X/XO. (B) EPR spectroscopy confirmed that superoxide is
generated by xanthine oxidase, and it can be distinguished from hydroxyl radical (C) with the spin trap DEPMPO. (D) Carrying out the reaction in the presence of 30 U/mL SOD
eliminated the spectra. The presence of P1 (E) and P2 (F), borane-deficient versions of PB1 and PB2 respectively, convert the spectra to that of the hydroxyl adduct of DEPMPO,
consistent with a reduction mechanism. Hydroethidine results are representative of 3 independent experiments. EPR spectroscopy results are representative of two independent
experiments.
Fig. 5. PB1 and PB2 are stable under accelerated stability conditions. Peak areas for PB1 at 40 ꢁC and 60 ꢁC were 80% and 72% of control respectively at 37 days. PB2 peak areas
were 100% of control for both conditions at 37 days. At 125 days, the peak area for PB1 at 40 ꢁC was 58% and PB2 was 86%; at 60 ꢁC peak areas for PB1 and PB2 were 12% and 14%
respectively. Results are from 1 experiment.

30
N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
PB1 and PB2 met the requirements of Lipinksi’s Rule of 5 for
druglikeness (Ghose et al., 1999; Lipinski et al., 2001): having fewer
than 5 hydrogen bond donors, fewer than 10 hydrogen bond ac-
ceptors, a log Pe between ꢂ0.4 and 5.6, and a molecular weight
between 160 and 480.
reduction of critical disulfides formed as a result of oxidation by
superoxide or other reactive oxygen species.
4.2. Upstream vs. downstream therapies
A critical issue with respect to cytoprotective therapies is the
timing between the onset of the injury, the commitment point for
dying (i.e. “point of no return”), and the point at which cell death
becomes manifest. If a drug only becomes bioavailable after the
commitment point, then it will not be cytoprotective, even if
administered long before cell death is manifest. This is analogous to
the fuse of a bomb, which will continue to burn even if the match
that ignited it is extinguished. In the case of cytoprotective thera-
pies, those that act on proximal or upstream pathways in cell death
need to be given early to be effective. Those that act on distal or
downstream pathways in cell death can be given later. The relation
between injury and the timing of cytoprotection is critical to clin-
ical trial design (Levin, 2004).
The neuroprotective action of phosphine-borane complexes is
associated with activation of ERK1/2 (Almasieh et al., 2011), a
downstream mechanism that also underlies the neuroprotection
mediated by brain-derived neurotrophic factor and the TrkB re-
ceptor (Cheng et al., 2002). Activation of this pathway with PB1
protects both the somal and axonal compartment after axonal
injury in two different animal models (Almasieh et al., 2011). The
intermediate steps between the reduction of a disulfide(s) and the
activation of the ERK1/2 pathway are not known, and presumably
reflect interruption of a cascade of redox signaling in one or more
critical death pathways, initiated by superoxide-dependent disul-
fide oxidation of key protein(s). Although c-Jun N-terminal kinase
(JNK) signaling is critical to RGC death after axonal injury
(Fernandes et al., 2012), the lack of effect of PB1 on phospho-JNK1/
2/3 levels indicates that its prosurvival effect bypasses JNK
(Almasieh et al., 2011).
The paradigm of upstream and downstream therapies can be
applied to the case of axonal injury in the CNS (Fig. 6). Our previous
work demonstrated that optic nerve transection activated an
intracellular pathway which caused elevated levels of superoxide
within the soma of the RGC (Kanamori et al., 2010a; Lieven et al.,
2006). This superoxide-dependent pathway is similar to that seen
in growth factor-deprived sympathetic neurons (Greenlund et al.,
1995) and renal proximal tubular cells (Lieberthal et al., 1998). It
likely represents an upstream pathway for cell death because it
precedes the appearance of cell death markers such as phosphati-
dylserine externalization (Kanamori et al., 2010a) and cytochrome c
release (Lieven et al., 2012). However, cell-permeant superoxide
scavengers such as pegylated superoxide dismutase (Kanamori
4. Discussion
4.1. Mechanism of action
These results provide a mechanism by which phosphine-borane
complexes are able to protect CNS cells from injury. Use of a fluo-
rescent dithiol reporter demonstrate that this class of drugs is able
to enter cells and reduce intracellular disulfides. CNS-specific
PAMPA-BBB assays showed that they penetrate brain lipids,
consistent with their observed neuroprotective potency in vivo
(Almasieh et al., 2011). Finally, although phosphine-borane com-
plexes inhibit a pathway for neuronal death that is superoxide-
dependent (Kanamori et al., 2010a; Lieven et al., 2006; Scott
et al., 2010), EPR and fluorescent reporter experiments indicate
that they are not superoxide scavengers but rather disulfide
reducing agents.
Previous studies demonstrated that axonal injury is associated
with a superoxide burst within the RGC soma (Kanamori et al.,
2010a; Lieven et al., 2006). Consistent with those findings, knock-
down of the endogenous mitochondrial superoxide scavenger su-
peroxide dismutase-2 increases neuronal cell death (Scott et al.,
2010). In the present study PB1 and PB2 had no superoxide scav-
enging activity, as manifested by absence of decrease in superoxide
produced in a xanthine/xanthine oxidase system and detected by
2-hydroxyethidium fluorescence or electron paramagnetic reso-
nance. The partial rescue of retinal ganglion cell death with PB1 and
PB2 seen in in vitro (Schlieve et al., 2006) and in vivo (Almasieh
et al., 2011) is therefore likely the result of restoration of redox
homeostasis by disulfide reduction and not superoxide scavenging.
Given that phosphine-borane complexes inhibit neuronal death
induced by elevations of superoxide in mitochondria external to the
matrix (Seidler et al., 2010), yet do not scavenge superoxide, this
implies that their disulfide-reducing activity is downstream of su-
peroxide generation.
Phosphines reduce protein disulfides through nucleophilic
attack by the reactive lone-pair electrons of the phosphorous.
Following hydrolysis this results first in a thiophosphonium salt
and then a free thiol and oxidized phosphine (Cline et al., 2004).
PB1 and PB2 were equally or more effective at reducing the disul-
fide bond of DTNB at pH 7 than equimolar concentrations of
commonly used reducing agents DTT and TCEP. PB1 and PB2 were
also able to reduce intracellular disulfides on a shorter time scale
than TCEP in tissue culture. P1 and P2, similar to PB1 and PB2 but
without borane protection, had strong reducing activity by EPR
spectroscopy, converting the superoxide adduct of the spin trap to a
hydroxyl adduct.
The increased effectiveness of PB1 and PB2 compared to the
parent compound TCEP results from specific structural modifica-
tions. Borane protection of the reactive phosphine protects the
compounds from extracellular oxidation. Intracellular phosphine
reactivity is restored when the borane protecting group is removed
by tertiary amines within tissues. Replacement of TCEP’s organic
acid groups by methyl esters results in intracellular sequestration
upon polarization of the molecule due to cleavage by esterases. In
addition, replacement of organic acid groups by phenyl rings re-
sults in increased cell membrane permeability and lower steric
hindrance of the lone-pair phosphorous electrons. Overall, these
modifications result in their concentration within cells where they
can interact with accessible protein thiols. This allows effective
et
al.,
2010a)
or
iron(III)
2,17-bis-sulfonato-5,10,15-
tris(pentafluorophenyl)corrole (Fe(tpfc)(SO(3)H)(2)) (Fe-corrole)
(Catrinescu et al., 2012) only incompletely protect RGCs against
axonal injury.
The present study provides a mechanism by which phosphine-
borane complex drugs would be more efficacious than superoxide
scavenging, based on the fact that they chemically reduce protein
disulfides which form under oxidative conditions (Fig. 6). This
redox chemistry is downstream of superoxide and other reactive
oxygen species, and therefore its reversal would be effective even if
late in the course of neuronal pathophysiology. Furthermore, if the
reactivity of reactive oxygen species with sulfhydryl groups on
critical signaling molecules is rapid, then this would help explain
why clinical trials based on scavenging superoxide and other
oxidative molecules fail to demonstrate efficacy (Louwerse et al.,
1995; Shuaib et al., 2007; Parkinson Study Group, 1993). It is un-
known whether therapies such as phosphine-borane complexes
are sufficiently downstream to be efficacious in randomized clinical

N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
31
for the latter alternative is the mechanism by which the WldS
mutation prevents axonal degeneration. These animals have a
fusion of nicotinamide mononucleotide adenylyl transferase 1
(NMNAT-1) to ubiquitination factor e4b (UBE4B) by an 18 amino
acid linkage. The mechanism of WldS axoprotection involves
increased axonal NMNAT-1 after axotomy (Babetto et al., 2010).
NMNAT-1 is critical to synthesis of NADþ, a key electron-accepting
redox agent, and it is possible that one of the redox reactions by
which elevated local NADþ protects axons interacts with
phosphine-borane complex redox chemistry.
In summary, phosphine-borane complexes interfere with
injury-mediated signals by reducing intracellular disulfide(s) and
not by superoxide scavenging. Pharmacological characterization of
these molecules is consistent with their being candidates for use in
the clinic based on their mechanism of action in optic nerve dis-
ease, lipophilic character, blood-brain barrier permeability, and
accelerated stability. These drugs may have efficacy and utility for
treating diseases where there is upstream superoxide signaling of
cell death.
Acknowledgments
We thank Dr. Christopher Felix, PhD and Jacek Zielonka, PhD of
the Clinical & Translational Science Institute of Southeast Wiscon-
sin, for assistance with EPR studies.
This work was supported by the National Institutes of Health
(Grants R21 EY017970, R21 EY025074, P30 EY016665); the Bright-
Focus Foundation; the Retina Research Foundation; and an unre-
stricted departmental grant from Research to Prevent Blindness.
The Clinical & Translational Science Institute of Southeast Wis-
consin is supported by the National Center for Research Resources
and the National Center for Advancing Translational Sciences
through National Institutes of Health (Grant UL1 RR031973). LAL is
Canada Research Chair in Translational Visual Science.
Fig. 6. Schematic for theoretical pathway linking neuronal injury with cell death
via superoxide and formation of disulfide bonds. The effect of phosphine-borane
complexes on the signaling of retinal ganglion cell death after axonal injury is distal
to an oxidative burst, with axonal injury inducing a burst of superoxide within the
retinal ganglion cell (RGC) soma (Kanamori et al., 2010a; Lieven et al., 2006). The time
course for superoxide induction is slow, and even a small amount of residual super-
oxide after scavenging will induce downstream oxidative disulfide formation and
subsequent neuronal death. Inhibiting downstream signaling events with phosphine-
borane complexes (Almasieh et al., 2011) reduces cell death by reducing critical
disulfides and indirectly inducing activation of prosurvival ERK1/2 pathways.
References
trials, although their ability to chemically reduce protein disulfides
and protect neurons in animal models of traumatic optic neurop-
athy and glaucoma is encouraging (Almasieh et al., 2011). It is also
possible that combined upstream and downstream therapies (e.g. a
superoxide scavenger and a phosphine-borane complex reducing
agent) would be more effective than either alone.
Al-Ubaidi, M.R., 2014. RGC-5: are they really 661W? the saga continues. Exp. Eye
Res. 119, 115.
Almasieh, M., Lieven, C.J., Levin, L.A., Di Polo, A., 2011. A cell-permeable phosphine-
borane complex delays retinal ganglion cell death after axonal injury through
activation of the pro-survival extracellular signal-regulated kinases 1/2
pathway. J. Neurochem. 118, 1075e1086.
Babetto, E., Beirowski, B., Janeckova, L., Brown, R., Gilley, J., Thomson, D.,
Ribchester, R.R., Coleman, M.P., 2010. Targeting NMNAT1 to axons and synapses
transforms its neuroprotective potency in vivo. J. Neurosci. 30, 13291e13304.
Beirowski, B., Babetto, E., Coleman, M.P., Martin, K.R., 2008. The WldS gene delays
axonal but not somatic degeneration in a rat glaucoma model. Eur. J. Neurosci.
28, 1166e1179.
Brisset, H., Gourdel, Y., Pellon, P., Le Corre, M., 1993. Phosphine-borane complexes;
direct use in asymmetric catalysis. Tetrahedron Lett. 34, 4523e4526.
Burns, J.A., Butler, J.C., Moran, J., Whitesides, G.M., 1991. Selective reduction of
disulfides by tris(2-carboxyethyl)phosphine. J. Org. Chem. 56, 2648e2650.
Cai, J., Jones, D.P., 1998. Superoxide in apoptosis. Mitochondrial generation triggered
by cytochrome c loss. J. Biol. Chem. 273, 11401e11404.
Carugo, O., Cemazar, M., Zahariev, S., Hudaky, I., Gaspari, Z., Perczel, A., Pongor, S.,
2003. Vicinal disulfide turns. Protein Eng. 16, 637e639.
Catrinescu, M.M., Chan, W., Mahammed, A., Gross, Z., Levin, L.A., 2012. Superoxide
signaling and cell death in retinal ganglion cell axotomy: effects of metal-
locorroles. Exp. Eye Res. 97, 31e35.
Cheng, L., Sapieha, P., Kittlerova, P., Hauswirth, W.W., Di Polo, A., 2002. TrkB gene
transfer protects retinal ganglion cells from axotomy-induced death in vivo.
J. Neurosci. 22, 3977e3986.
Cline, D.J., Redding, S.E., Brohawn, S.G., Psathas, J.N., Schneider, J.P., Thorpe, C., 2004.
New water-soluble phosphines as reductants of peptide and protein disulfide
bonds: reactivity and membrane permeability. Biochemistry 43, 15195e15203.
Di, L., Kerns, E.H., Fan, K., McConnell, O.J., Carter, G.T., 2003. High throughput arti-
ficial membrane permeability assay for blood-brain barrier. Eur. J. Med. Chem.
38, 223e232.
Fernandes, K.A., Harder, J.M., Fornarola, L.B., Freeman, R.S., Clark, A.F., Pang, I.H.,
John, S.W., Libby, R.T., 2012. JNK2 and JNK3 are major regulators of axonal
injury-induced retinal ganglion cell death. Neurobiol. Dis. 46, 393e401.
Geiger, L.K., Kortuem, K.R., Alexejun, C., Levin, L.A., 2002. Reduced redox state
4.3. Neuroprotection vs. axoprotection
Therapies for primary axonal diseases are necessarily different
from diseases where the initial site of injury is the cell body because
protection of the cell body may not be sufficient to rescue the
injured axon. This issue of somaprotection vs. axoprotection has
been reviewed in the past (Levin, 2001; Schwartz et al., 1999), and
is relevant to the discovery of effective neuroprotective therapies.
For example, it is possible to completely block RGC death by
knocking out bax, resulting in preservation of the cell body in
experimental glaucoma yet not preventing axonal degeneration
(Howell et al., 2007; Libby et al., 2005). Conversely, slowing down
axonal degeneration in the Wallerian degeneration slow (WldS) rat
does not prevent loss of the cell body in experimental glaucoma
(Beirowski et al., 2008).
In the case of phosphine-borane complexes, PB1 delay both cell
body and axonal loss in optic nerve transection and experimental
glaucoma (Almasieh et al., 2011). This indicates that the disulfide-
reduction mechanism of action either inhibits the early pathology
at the site of injury (the axon) and/or interrupts a subsequent
signaling process for axonal and cell body degeneration. Support

32
N.J. Niemuth et al. / Neurochemistry International 99 (2016) 24e32
allows prolonged survival of axotomized neonatal retinal ganglion cells.
Neuroscience 109, 635e642.
Ghose, A.K., Viswanadhan, V.N., Wendoloski, J.J., 1999. A knowledge-based approach
in designing combinatorial or medicinal chemistry libraries for drug discovery.
1. A qualitative and quantitative characterization of known drug databases.
J. Comb. Chem. 1, 55e68.
Greenlund, L.J., Deckwerth, T.L., Johnson, E.M., 1995. Superoxide dismutase delays
neuronal apoptosis: a role for reactive oxygen species in programmed neuronal
death. Neuron 14, 303e315.
release in a growth factor deprivation model. Apoptosis 17, 591e599.
Lipinski, C.A., Lombardo, F., Dominy, B.W., Feeney, P.J., 2001. Experimental and
computational approaches to estimate solubility and permeability in drug
discovery and development settings. Adv. Drug Deliv. Rev. 46, 3e26.
Louwerse, E.S., Weverling, G.J., Bossuyt, P.M., Meyjes, F.E., de Jong, J.M., 1995. Ran-
domized, double-blind, controlled trial of acetylcysteine in amyotrophic lateral
sclerosis. Arch. Neurol. 52, 559e564.
Park, C., Raines, R.T., 2001. Adjacent cysteine residues as a redox switch. Protein Eng.
14, 939e942.
Hampton Research, 2016. Material Safety Data Sheet for Tris(2-carboxyethyl)
Phosphine Hydrochloride. Accessed April 26, 2016. http://hamptonresearch.
com/documents/product/hr007314_2-651_sds.pdf.
Howell, G.R., Libby, R.T., Jakobs, T.C., Smith, R.S., Phalan, F.C., Barter, J.W., Barbay, J.M.,
Marchant, J.K., Mahesh, N., Porciatti, V., Whitmore, A.V., Masland, R.H.,
John, S.W., 2007. Axons of retinal ganglion cells are insulted in the optic nerve
early in DBA/2J glaucoma. J. Cell Biol. 179, 1523e1537.
Kanamori, A., Catrinescu, M.M., Kanamori, N., Mears, K.A., Beaubien, R., Levin, L.A.,
2010a. Superoxide is an associated signal for apoptosis in axonal injury. Brain
133, 2612e2625.
Kanamori, A., Catrinescu, M.M., Mahammed, A., Gross, Z., Levin, L.A., 2010b. Neu-
roprotection against superoxide anion radical by metallocorroles in cellular and
murine models of optic neuropathy. J. Neurochem. 114, 488e498.
Karoui, H., Hogg, N., Frejaville, C., Tordo, P., Kalyanaraman, B., 1996. Characterization
of sulfur-centered radical intermediates formed during the oxidation of thiols
and sulfite by peroxynitrite. ESR-spin trapping and oxygen uptake studies.
J. Biol. Chem. 271, 6000e6009.
Krishnamoorthy, R.R., Clark, A.F., Daudt, D., Vishwanatha, J.K., Yorio, T., 2013.
A forensic path to RGC-5 cell line identification: lessons learned. Invest Oph-
thalmol. Vis. Sci. 54, 5712e5719.
Levin, L.A., 2001. Relevance of the site of injury of glaucoma to neuroprotective
strategies. Surv. Ophthalmol. 45, S243eS249.
Levin, L.A., 2004. Extrapolation of animal models of optic nerve injury to clinical
trial design. J. Glaucoma 13, 1e5.
Libby, R.T., Li, Y., Savinova, O.V., Barter, J., Smith, R.S., Nickells, R.W., John, S.W., 2005.
Susceptibility to neurodegeneration in a glaucoma is modified by Bax gene
dosage. PLoS Genet. 1, 17e26.
Parkinson Study Group, 1993. Effects of tocopherol and deprenyl on the progression
of disability in early Parkinson’s disease. N. Engl. J. Med. 328, 176e183.
Pullela, P.K., Chiku, T., Carvan 3rd, M.J., Sem, D.S., 2006. Fluorescence-based
detection of thiols in vitro and in vivo using dithiol probes. Anal. Biochem. 352,
265e273.
Schlieve, C.R., Tam, A., Nilsson, B.L., Lieven, C.J., Raines, R.T., Levin, L.A., 2006. Syn-
thesis and characterization of a novel class of reducing agents that are highly
neuroprotective for retinal ganglion cells. Exp. Eye Res. 83, 1252e1259.
Schwartz, M., Yoles, E., Levin, L.A., 1999. ’Axogenic’ and ’somagenic’ neurodegen-
erative diseases: definitions and therapeutic implications. Mol. Med. Today 5,
470e473.
Scott, C.J., Seidler, E.A., Levin, L.A., 2010. Cell-autonomous generation of mito-
chondrial superoxide is a signal for cell death in differentiated neuronal pre-
cursor cells. Brain Res. 1306, 142e148.
Seidler, E.A., Lieven, C.J., Thompson, A.F., Levin, L.A., 2010. Effectiveness of novel
borane-phosphine complexes in inhibiting cell death depends on the source of
superoxide production induced by blockade of mitochondrial electron trans-
port. ACS Chem. Neurosci. 1, 95e103.
Shuaib, A., Lees, K.R., Lyden, P., Grotta, J., Davalos, A., Davis, S.M., Diener, H.C.,
Ashwood, T., Wasiewski, W.W., Emeribe, U., 2007. NXY-059 for the treatment of
acute ischemic stroke. N. Eng. J. Med. 357, 562e571.
Simon, H.U., Haj-Yehia, A., Levi-Schaffer, F., 2000. Role of reactive oxygen species
(ROS) in apoptosis induction. Apoptosis 5, 415e418.
Sohn, Y.T., Park, B.Y., 2003. Characterization of the physicochemical properties of
KR-31378. Arch. Pharm. Res. 26, 526e531.
Swanson, K.I., Schlieve, C.R., Lieven, C.J., Levin, L.A., 2005. Neuroprotective effect of
sulfhydryl reduction in a rat optic nerve crush model. Invest Ophthalmol. Vis.
Sci. 46, 3737e3741.
Lieberthal, W., Triaca, V., Koh, J.S., Pagano, P.J., Levine, J.S., 1998. Role of superoxide
in apoptosis induced by growth factor withdrawal. Am. J. Physiol. 275,
F691eF702.
Lieven, C.J., Schlieve, C.R., Hoegger, M.J., Levin, L.A., 2006. Retinal ganglion cell
axotomy induces an increase in intracellular superoxide anion. Invest Oph-
thalmol. Vis. Sci. 47, 1477e1485.
Thompson, A.F., Crowe, M.E., Lieven, C.J., Levin, L.A., 2015. Induction of neuronal
morphology in the 661W cone photoreceptor cell line with staurosporine. PLoS
One 10, e0145270.
Zhao, H., Joseph, J., Fales, H.M., Sokoloski, E.A., Levine, R.L., Vasquez-Vivar, J.,
Kalyanaraman, B., 2005. Detection and characterization of the product of
hydroethidine and intracellular superoxide by HPLC and limitations of fluo-
rescence. Proc. Natl. Acad. Sci. U. S. A. 102, 5727e5732.
Lieven, C.J., Thurber, K.A., Levin, E.J., Levin, L.A., 2012. Ordering of neuronal
apoptosis signaling: a superoxide burst precedes mitochondrial cytochrome c