The Degradation Chemistry of GSK2879552: Salt Selection and Microenvironmental pH Modulation to Stabilize a Cyclopropyl Amine

Article abstract of DOI:10.1016/j.xphs.2019.04.026

The cyclopropyl amine moiety in GSK2879552 (1) degrades hydrolytically in high pH conditions. This degradation pathway was observed during long-term stability studies and impacted the shelf life of the drug product. This article describes the work to identify the degradation impurities, elucidate the degradation mechanism, and design a stable drug product. It was found that salt selection and control of the microenvironmental pH of the drug product formulation blend significantly improved the chemical stability of the molecule in the solid state.

Full text of DOI:10.1016/j.xphs.2019.04.026

Journal of Pharmaceutical Sciences xxx (2019) 1-7
Contents lists available at ScienceDirect
Journal of Pharmaceutical Sciences
journal homepage: www.jpharmsci.org
Drug DiscoveryeDevelopment Interface
The Degradation Chemistry of GSK2879552: Salt Selection and
Microenvironmental pH Modulation to Stabilize a Cyclopropyl
Amine
John M. Campbell 1 , Mei Lee 2 , Jacalyn Clawson , Sonya Kennedy-Gabb ,
,
*
,
*
1
1
3
1
2
1
Sarah Bethune , Ashley Janiga , Leanda Kindon , Kevin P. Leach
Analytical Sciences and Development, GlaxoSmithKline, Upper Providence, Pennsylvania
Product and Process Engineering, GlaxoSmithKline, Stevenage, Hertfordshire, UK
Drug Product Design and Development, GlaxoSmithKline, Upper Providence, Pennsylvania
1
2
3
a r t i c l e i n f o
a b s t r a c t
Article history:
The cyclopropyl amine moiety in GSK2879552 (1) degrades hydrolytically in high pH conditions. This
degradation pathway was observed during long-term stability studies and impacted the shelf life of the
drug product. This article describes the work to identify the degradation impurities, elucidate the
degradation mechanism, and design a stable drug product. It was found that salt selection and control of
the microenvironmental pH of the drug product formulation blend significantly improved the chemical
Received 6 December 2018
Revised 25 March 2019
Accepted 23 April 2019
stability of the molecule in the solid state.
Keywords:
chemical stability
salt selection
®
©
2019 American Pharmacists Association . Published by Elsevier Inc. All rights reserved.
forced conditions
formulation
physical characterization
Introduction
was unstable in solution at high pH. The principal route of degra-
dation observed during stress testing is shown in Figure 1.
GSK2879552 (1) is an orally bioavailable lysine-specific deme-
The susceptibility of 1 toward degradation at high pH was se-
thylase 1 inhibitor which was in clinical development for the po-
tential treatment of cancer. During drug development, routine
vere, with complete degradation of the drug substance occurring in
0.1 M NaOH after just 2 hours at 80 C. Rapid degradation of the
1
ꢀ
stress testing studies were conducted to support validation of an
HPLC impurities method. The studies revealed that the compound
molecule was still observed even at reduced temperatures. The
degradation byproduct(s) of primary amine 2 could not be identi-
fied during the initial stress testing study.
The instability observed in the stress testing study fore-
shadowed instability in the drug product (active pharmaceutical
ingredient [API] powder in a capsule for oral administration).
Indeed, the formation of 2 was observed during International
Council for Harmonisation of Technical Requirements for
Pharmaceuticals for Human Use-compliant stability studies. As
a result, the drug product required refrigerated storage to achieve
a shelf life sufficient to support clinical trials. While this was
suitable to progress clinical trials, neither the drug product
configuration, shelf life, nor storage condition were ideal for a
commercial product. Improvements to the drug product hinged
on understanding and controlling the chemical stability of 1.
Here, we present the investigational work performed to un-
derstand the solution-phase chemical degradation pathway and
offer mechanisms to explain its occurrence in the solid drug
Abbreviations used: API, active pharmaceutical ingredient; RH, relative humid-
ity; XRPD, X-ray powder diffraction; DAD, diode array detection; DSC, differential
scanning calorimetry; DVS, dynamic vapor sorption; HPLC, high performance liquid
chromatography; ICH, International Council for Harmonisation of Technical Re-
quirements for Pharmaceuticals for Human Use; NMR, nuclear magnetic resonance;
NT, not tested; TGA, thermogravimetric analysis.
Campbell and Lee contributed equally to this work.
Current address for Dr. Jacalyn Clawson: Janssen, Pharmaceutical Companies of
Johnson and Johnson, 1400 McKean Road, Spring House, Pennsylvania, 19477.
Current address for Dr. Sarah Bethune: Cambrex Corporation, 2620 Trade Centre
Avenue, Longmont, Colorado, 80503.
This article contains supplementary material available from the authors by request
or via the Internet at https://doi.org/10.1016/j.xphs.2019.04.026.
*
Correspondence to: John M. Campbell (Telephone: þ1-484-923-3441) and Mei
Lee (Telephone: þ44-143-876-2115).
E-mail addresses: john.m.campbell@gsk.com (J.M. Campbell), mei.2.lee@gsk.
com (M. Lee).
https://doi.org/10.1016/j.xphs.2019.04.026
®
0
022-3549/© 2019 American Pharmacists Association . Published by Elsevier Inc. All rights reserved.

2
J.M. Campbell et al. / Journal of Pharmaceutical Sciences xxx (2019) 1-7
Figure 1. Structures for the principle degradation impurities of GSK2879552 (1).
product. We used those mechanistic considerations to propose
strategies to improve the chemical stability of 1, including selection
of a less soluble salt form and microenvironmental pH modulation.
This was achieved by dissolving the dihydrochloride in
aqueous base and extracting the product with an organic solvent.
In this case, potassium carbonate and 10% methanol:dichloro-
methane was used. The extract was dried with magnesium sul-
fate to remove excess water and was then filtered. The filtrate
was evaporated under reduced pressure to yield amorphous
freebase.
To conduct the salt screen, the freebase was combined with
selected counterions in 4 different solvents and then subjected to
thermal cycling to encourage the formation of crystalline salts.
The counterions explored included those in the sulfonic acid
family such as p-toluenesulfonic acid, benzene sulfonic acid,
methanesulfonic acid, and 1,5-naphthalenedisulfonic acid; oxalic
acid, citric acid, phosphoric acid, and pamoic acid were also
evaluated.
Experimental
Materials and Methods
GSK2879552 (1) (GSK) is 4-((4-((((1R,2S)-2-phenylcyclopropyl)
amino)methyl)piperidin-1-yl)methyl)benzoic acid and is manu-
factured as the hydrochloride dihydrate salt. The pK ’s of 1 were
a
experimentally determined to be 9.04, 7.21, and 3.21 by UV-metric
ꢀ
titration in methanol-water at 25 C.
All other reagents and solvents were purchased from commer-
cial suppliers.
The solid products obtained from the screen were initially
examined by polarized light microscopy to confirm crystallinity
prior to further characterization by X-ray powder diffraction
Forced Degradation Methodology
(XRPD). The solids were also evaluated for hygroscopicity by
All stress testing conducted during mechanism elucidation was
performed in solution using the dihydrochloride salt of 1. Subse-
quently prepared salts and formulation blends were stressed in the
solid state.
gravimetric vapor sorption and physical stability. Physical stability
was assessed by XRPD analysis after gravimetric vapor sorption and
thermal cycling via variable temperature and variable humidity
XRPD.
High pH stress testing of 1 was performed by dissolving
the compound in 0.1 M sodium hydroxide at a concentration of
Aqueous Solubility and Saturated Aqueous pH Measurement
1.0 mg/mL. The solution was sealed air tight in a serum vial and
stored at the desired temperature, protected from light. After
stressing, the sample was equilibrated to room temperature,
quenched with an equivalent amount of 0.1 M hydrochloric acid,
and diluted for HPLC analysis to 0.1 mg/mL with the diluent
described in the analytical method.
To determine the solubility of the salts, an excess amount of
ꢀ
ꢀ
solid material was added to water at 25 C and 50 C to obtain a
slurry. The concentration of 1 in the supernatant was determined
after 24 h by HPLC. The pH of the supernatant was measured with a
pH meter.
Stress testing in sodium carbonate was conducted by dissolving
1
in 4 mM sodium carbonate at a concentration of 1.0 mg/mL. The
ꢀ
resulting solution was stored at 40 C for 4 days and quenched with
Analytical Methods
0
.1 M hydrochloric acid. These conditions led to ~75% degradation
of and minimized secondary degradation of hydro-
1
All impurities testing was performed using the HPLC method
described in Table 1.
cinnamaldehyde (3), which was extracted from the acidified
stressed solution in high purity with dichloromethane. The organic
extract was dried over magnesium sulfate, filtered, and concen-
trated under reduced pressure to give 3 as a clear, colorless oil. NB:
compound 3 degrades slowly in the presence of atmospheric oxy-
gen to give hydrocinnamic acid.
Diffraction patterns were collected using a Panalytical X’Pert Pro
diffractometer with Cu K
a
radiation, with a generator voltage and
current of 40 kV and 40 mA, respectively. All diffraction patterns
ꢀ
ꢀ
ꢀ
were collected from 2 -40 2theta with a 0.02 2theta scan step and
a step time of 0.1 s. Variable temperature XRPD patterns were
Solid state stress testing was conducted by placing the material
in a serum vial, sealing air tight, and storing the vial in a
temperature-regulated oven for the duration of stressing. If desired,
relative humidity (RH) was controlled by placing the material to be
stressed in an unsealed vial within a glass micro-desiccator con-
ꢀ
ꢀ
ꢀ
collected by heating from 25 C to 150 C at 10 C/min with holds at
0 C, 80 C, and 150 C before cooling back to 25 C.
ꢀ
ꢀ
ꢀ
ꢀ
5
Hygroscopicity was measured by vapor sorption using a DVS
ꢀ
Advantage or DVS-1. Samples were analyzed at 25 C over a relative
humidity range from 0% to 90% in 10% RH increments, as well as at
2
taining a saturated aqueous solution of appropriate salt and sealed
75% RH. The RH was increased sequentially with step adjustments
with a plastic screw lid. The micro-desiccator containing the sam-
ple was then stored in a temperature-regulated oven for the
duration of stressing.
made after the rate of mass change (dm/dt) was <0.02% or after a
hold time of 3 hours was reached, whichever came first.
Differential scanning calorimetry was performed on a TA In-
struments Discovery series differential scanning calorimetry in
open aluminum pans. Thermogravimetric analysis was performed
on a TA Instruments Discovery series thermogravimetric analysis
Salt Screening Methodology
ꢀ
To perform a salt screen, the dihydrochloride salt of 1 was
initially converted to the freebase.
in platinum crucibles. Samples were heated to 250 C at a rate of
ꢀ
10 C/min.

J.M. Campbell et al. / Journal of Pharmaceutical Sciences xxx (2019) 1-7
3
Table 1
HPLC Impurities Method for GSK2879552 (1)
Instrument Details
Agilent 1100 Series HPLC Equipped with Diode Array Detector
Mobile phase A
Mobile phase B
Flow rate
Water, 0.05% (v/v) trifluoroacetic acid
Acetonitrile, 0.05% (v/v) trifluoroacetic acid
1 mL/min
Injection volume
Nominal sample concentration
Sample diluent
Column details
Column temperature
Gradient profile
20 mL
0.1 mg/mL
Water:acetonitrile, (4:1, v/v)
Atlantis dC18 (3.9 ꢁ 150 mm, 3
mm particle diameter)
ꢀ
40 C
Time (min)
% mobile phase A
% mobile phase B
0
2
3
3
3
.00
95
60
5
95
95
5
0.00
0.00
0.01
8.00
40
95
5
5
a
Detection wavelength
230 nm, 4 nm bandwidth, 4 nm slit width, reference wavelength off
a
A detection wavelength of 215 nm was used to detect the presence of compound 3.
Results and Discussion
Subsequent work demonstrated that 3 itself was not stable in
the stress testing media. It degraded rapidly in sodium hydroxide
via a bimolecular aldol condensation. In hindsight, secondary
degradation, poor UV absorbance at the analytical wavelength, and
poor ionization by most common mass spectrometry techniques all
compounded efforts to detect 3 in stressed samples.
Stress Testing Experiments to Elucidate the Chemical Degradation
Mechanism
We began our work by trying to understand the chemical
mechanism of degradation. Because the degradation byproduct(s)
of 2 were not identified during the initial stress testing study, the
fate of the cyclopropyl moiety in 1 became the first objective of our
inquiry.
After a careful evaluation of the possible degradation mecha-
nisms, hydrocinnamaldehyde was proposed as a potential degra-
dation byproduct of 2. As this compound is commercially available,
a marker was purchased and analyzed by the analytical method
with diode array detection. As shown in Figure 2, the UV absor-
bance of hydrocinnamaldehyde (3) is negligible at the analytical
wavelength (230 nm), demonstrating that the analytical method
was not suitable for direct detection of the aldehyde. However,
adjustment of the analytical wavelength and comparison of the
stressed sample chromatograms with that of the purchased marker
revealed the unambiguous presence of 3 in the base-stressed
samples.
When 1 is stressed under gentler conditions by substituting
sodium carbonate for sodium hydroxide, clean conversion of the
drug substance to amine 2 and aldehyde 3 was readily achieved
with minimal evidence of secondary degradation of 3. From such a
reaction, 3 could be extracted in high purity with dichloromethane
for structural confirmation by nuclear magnetic resonance.
To interrogate the mechanism of degradation further, a labeling
experiment was performed by stressing the drug substance with
sodium carbonate in deuterium oxide. After purification of 3 from
1
the stress testing reaction as before, H-nuclear magnetic reso-
nance analysis revealed deuteration of 3 at the locations indicated
in Figure 3. This information, in addition to the clear dependence of
the reaction rate on pH, was used to propose a mechanism for the
degradation reaction.
We propose that in a high pH environment, the secondary
amine in 1 is deprotonated and the electron-donating lone pair of
electrons on the nitrogen facilitates ring opening of the cyclo-
a
propane. Based on the high pK ’s of benzylic carbons, protonation
at the benzylic position by any donor present likely occurs
concomitant with ring opening to give imine 4. Release of the
strain energy of the cyclopropane likely drives the ring-opening
reaction, and it seems reasonable that the phenyl ring could sta-
bilize the transition state leading to 4 by hyperconjugation. Imine
4
was never directly observed but evidently undergoes hydrolysis
to form 2 and 3 as stable degradation products, occurring either
immediately after the ring opening or perhaps during analytical
sample preparation.
The chemistry depicted in Figure 3 is consistent with the known
3
ability of cyclopropanes to mimic the chemistry of olefins. Viewed
from this perspective, the amine-substituted cyclopropane may be
thought of as an enamine analog. The mechanism also represents
an example of classic donor acceptoresubstituted cyclopropane
chemistry, which has been explored thoroughly in the field of
4
-7
synthetic chemistry.
The use of amines as electron donors in
8
-11
such reactions are documented in the literature,
including ex-
amples of nitrogen-substituted cyclopropanes undergoing ring
opening and subsequent hydrolysis to yield aldehydes in a manner
Figure 2. UV spectrum of hydrocinnamaldehyde (3), determined by HPLC-DAD. The
detection wavelength for the analytical method was 230 nm. At this wavelength,
the analytical method was not suitable for detecting 3 but readily detected the pres-
ence of 2.
12,13
analogous to the present case.
Interestingly, while the literature
is replete with examples of aryl donor groups, to our knowledge the

4
J.M. Campbell et al. / Journal of Pharmaceutical Sciences xxx (2019) 1-7
Figure 3. Proposed mechanism for the degradation of GSK2879552 (1) at high pH. When stress testing is conducted in deuterium oxide, aldehyde 3 is deuterated at the positions
colored red. Deuteration of the alpha carbons in 3 likely arises from either keto-enol tautomerism of 3, imine-enamine tautomerism of 4, or both.
only examples of aryl acceptor groups also incorporated an amine
donor.
Alternatively, the apparent degradation of 1 observed in the
solid state might in actuality occur via the solution-phase mecha-
nism after dissolution of the API into any amount of sorbed mois-
14,15
Our mechanistic findings with respect to 1 are consistent with
observations made earlier in drug development: the freebase of 1
2
2
ture. Guerrieri et al.
showed that the solid-state chemical
ꢀ
degrades even when stored at ꢂ20 C, evidently caused by the
degradation of procaine salts was entirely consistent with the rates
of degradation measured in solution. Their findings suggesting that
degradation in water layered onto the surface of the solid state
particles is the primary source of chemical degradation observed in
the procaine salts. Following this mechanism, aqueous solubility,
quantity of sorbed moisture, and pH of the water layer (which is
presumably saturated with the API) play key roles in the solid-state
degradation of 1.
secondary amine being unprotonated. Thus, it was recognized early
in development that 1 needs to be salted to achieve reasonable
stability.
With the mechanistic information in hand, a more precise un-
derstanding of the impact of pH on the degradation reaction rate
was explored. 1 was stressed at 1.0 mg/mL in solutions of various
ꢀ
pH values at 40 C. HPLC analysis of these solutions over time
allowed first-order reaction rate constants to be determined (First
order kinetics was assumed. The experiments performed did not
unambiguously reveal a reaction order or rate law equation, but
rather were designed to estimate the relative rates of the reaction
as a function of pH. The assumption of a different reaction order
does not meaningfully alter the pH-rate plot.). Figure 4 shows a plot
of reaction rate versus pH with the data fit to a sigmoidal curve (See
the associated Supplementary Material for more details on kinetic
data fitting, including the determination of additional reaction
parameters.), as might be expected for a reaction whose rate is
Ultimately, we were unable to demonstrate unequivocally the
operative mechanism of degradation in the solid state. This was
due, in part, to the severe instability of the freebase of 1, which
made performing the necessary experiments challenging. Still, we
recognized that in either of the 2 mechanistic possibilities that we
16
dependent on the protonation state of the molecule. Additionally,
the ionization state distribution for 1 was calculated from the
experimentally determined pK ’s and is superimposed on Figure 4.
a
An examination of the correlation between the reaction rate and
the pH-dependent speciation of 1 further supported our mecha-
nistic proposal that deprotonation of the secondary amine leads to
the onset of the degradation reaction.
Extrapolating From the Solution Phase to the Solid State
After completion of the solution phase stress testing studies, we
turned our attention to the degradation observed in the drug prod-
uct. In extrapolating our mechanistic understanding to the solid
state, we considered 2 potential routes of chemical degradation.
First, any disproportionation of 1 will potentially lead to a
deprotonated secondary amine and trigger degradation. It is
generally believed that disproportionation of an API is a solution-
mediated process driven by the solubility of the API (both the salt
and freebase), microenvironmental pH, and, in drug products, the
17-20
properties of the excipients.
Hence, disproportionation of 1
would involve the dissolution of the salt into any moisture sorbed
onto the surface of the solid material followed by precipitation of
the freebase, which would then degrade in the solid state. In the
case of hydrochloride salts, the volatility of hydrochloric acid may
ꢀ
Figure 4. pH-rate profile for the degradation of GSK2879552 (1) at 40 C in water (dark
circles with dashed line), overlaid with a plot of the ionization states of 1 from pH 0-14.
The data support our hypothesis that deprotonation of the secondary amine in 1 leads
to chemical degradation. Each chemical structure is colored to correspond with its
respective mole fraction curve.
21
exacerbate this mechanism.

J.M. Campbell et al. / Journal of Pharmaceutical Sciences xxx (2019) 1-7
5
considered, reducing the solubility of 1 ought to improve its
chemical stability.
easily predicted. Hence, we selected several hydrophobic counter-
ions, conducted a salt screen of 1, and evaluated the physical and
chemical properties of the synthesized salts against the dihydro-
chloride and the criteria described above.
Thus far,1 had been developed as a dihydrochloride salt, and it is
this salt form that was evaluated during the original stress testing
and drug product formulation. The dihydrochloride is nonhygro-
scopic, which would minimize the amount of sorbed moisture. In
addition, the pH of a saturated solution of the dihydrochloride is
below the pH at which the degradation rate begins to accelerate
Salt Screen Results
A salt screen of 1 was conducted with 8 counterions. The results
are summarized in Table 2. Four of the counterions produced crys-
talline solids. The besylate and pamoate were found to be physically
unstable during physical characterization. The napadisylate and
ditosylate salts were physically stable, with vapor sorption and
thermal analysis behavior consistent with dihydrate crystalline
structures. Our intuition was to identify an anhydrous salt if
possible, as we wondered if an anhydrate could help minimize
degradation. However, no stable anhydrous salts were discovered.
Polymorph screens were conducted on the ditoslyate and
napadisylate salts, which revealed the ditosylate dihydrate and the
napadisylate dihydrate to be the most thermodynamically stable
forms. Single-crystal structures of both salts were also determined.
The chemical structures of these salts are shown in Figure 5.
The aqueous solubility and pH of the napadyslate and ditosylate
salts were determined experimentally as shown in Table 3. The data
confirmed that both salts met our predetermined acceptance
criteria. They are less soluble than the dihydrochloride, and the
aqueous pH of a saturated solution remains below 5.
(
see Fig. 4), which ought to minimize solution-phase degradation in
any moisture layer (Note that in the drug product, the API is in
contact with excipients which potentially overwhelm the API with
respect to determining the pH of any sorbed moisture. Thus, while
the saturated aqueous pH of the API salt is certainly an important
consideration, it may not be the best handle for controlling the
degradation reaction in the drug product.). However, the dihydro-
chloride is highly soluble in water (>48 mg/mL). Thus, regardless of
whether chemical degradation of 1 occurs in the water layer on the
surface of the particles or via disproportionation, selecting a salt
with reduced solubility seemed likely to improve the chemical
stability of the solid.
Based on these considerations, we established the following
criteria for selecting an alternative salt of 1:
1
. aqueous solubility <48 mg/mL but >0.12 mg/mL in physiological
conditions (A reduction in solubility can impact the bio-
pharmaceutics of the molecule. Therefore, we selected salts that
ꢀ
had solubilities at 37 C over the entire BCS pH range (1-7) and in
biorelevant media which were consistent with a highly soluble
compound (data not shown). This translates to a minimum
aqueous solubility of >0.12 mg/mL at 37 C for GSK2879552 (1).)
. saturated aqueous pH < 5 (Note that in the drug product, the API
is in contact with excipients which potentially overwhelm the
API with respect to determining the pH of any sorbed moisture.
Thus, while the saturated aqueous pH of the API salt is certainly
an important consideration, it may not be the best handle for
controlling the degradation reaction in the drug product.)
. nonhygroscopic or slightly hygroscopic
Stress Testing of Alternative Salts
ꢀ
In order to quickly gauge their relative chemical stability, the
ditosylate and napadisylate were evaluated in a stress testing study
(Table 4). Both salts showed an improvement in chemical stability
over the original dihydrochloride. No significant degradation was
2
ꢀ
observed for the samples stressed at 80 C and ambient humidity.
ꢀ
The degradation of 1 at 80 C and 75% RH correlated well with the
measured aqueous solubility, providing the first evidence sup-
porting our hypothesis that a decrease in solubility might improve
the chemical stability.
3
4
. physically stable with no evidence of disproportionation
We decided to carry the napadisylate forward into drug product
development as it presented the most promise in terms of chemical
stability, physical stability, reduced solubility, and minimal hygro-
scopicity. The drug substance stress testing results were
The aqueous solubility of a salt can be reduced by selecting
counterions that are more hydrophobic, although it should be
noted that solubility is dependent on several factors and thus not
2
3
Table 2
Salt Screen and Physical Characterization of GSK2879552 (1) Salts
Anticipated Relative Aqueous Solubility (mg/mL)a
Counterion
Crystallinity
Hygroscopicityb
Physical Stabilityc
1
0
0
Hydrochloride
Mesylate
Phosphate
Citrate
Besylate
Napadisylate
Tosylate
✓ (hemihydrate)
7
✓ (<0.1%)d
NT
✓
e
NT
NT
NT
ꢂ1
ꢂ2
1
7
NT
NT
7 (~4%)
7
f
f
✓ (hydrate)
✓ (dihydrate)
✓ (dihydrate)
7
g
10
✓ (<0.3%)
✓ (~1.5%)
✓
✓
g
h
Oxalate
Pamoate
7
NT
7 (>12%)
NT
7
ꢂ2
i
<
10
✓
a
b
c
Relative aqueous solubility anticipated based on hydrophobicity of the counterions and the solubilities reported for procaine salts.22
Weight gain from 10% to 80% RH in parentheses.
Physical stability indicated by no change in solid form observed after exposure to humidity cycling during gravimetric vapor sorption analysis, no change in solid form
ꢀ
ꢀ
observed up to 80 C during variable temperature (VT) X-ray powder diffraction (XRPD), and rehydration back to the same form after thermal cycling to 150 C (past
dehydration).
d
No dehydrate observed at <10% RH; hemihydrate confirmed by VT-XRPD and single-crystal X-ray diffraction.
Not tested.
Reversible dehydration observed at <10% RH; change in solid form observed after gravimetric vapor sorption and during VT-XRPD.
Reversible dehydration observed at <10% RH; weight loss below 10% consistent with dihydrate.
Evidence of anhydrate formation during drying, requiring exposure to high moisture (>ambient) to rehydrate.
Hygroscopicity is likely due to the formation of amorphous during the first sorption cycle.
e
f
g
h
i

6
J.M. Campbell et al. / Journal of Pharmaceutical Sciences xxx (2019) 1-7
Figure 5. Chemical structures of GSK2879552 (1) counterions selected for further evaluation. The dihydrochloride salt was the original salt.
Table 3
showed incompatibility with nearly all excipients evaluated (data
not shown).
If we extend our hypotheses around the solid-state degradation
mechanisms of 1 to the drug product, 2 things become apparent.
First, the hygroscopicity of the excipients is likely to be the domi-
nant cause of moisture uptake in the drug product. Second, the
acidity or basicity of the excipients could play a role in the degra-
dation by either facilitating disproportionation, modulating the pH
of any water on the surface of the solid particles, or both.
Aqueous Solubility and pH of GSK2879552 (1) Salts
Salt
pH (Saturated
Aqueous, 25 C)
Solubility
(mg/mL, 25 C)
Solubility
(mg/mL, 50 C)
ꢀ
ꢀ
ꢀ
Dihydrochloride
Ditosylate
Napadisylate
2.33
3.12
2.95
>48
1.3
0.1
NT
3.1
0.3
NT, not tested.
Despite the fact that excipient properties likely overwhelm the
API with respect to moisture sorption and pH modulation, we still
expected that reducing the solubility of 1 should improve its
chemical stability in the presence of excipients. In addition, we saw
in the drug product an opportunity to stabilize the molecule by
using acidic modifiers, a means of control that is not generally an
Table 4
Stress Testing of Selected GSK2879552 (1) Salts in the Solid State
Condition
Dihydrochloride
Ditosylate
Napadisylate
Initial
ND
ND
4.19
0.91
1.19 (0.28)
2.86 (1.95)
0.64
0.62 (ND)
0.76 (0.12)
ꢀ
ꢀ
a
8
8
0 C, ambient RH, 1 wk
29,30
option in the drug substance.
0 C, 75% RH, 1 wk
We set out to test these hypotheses by comparing the napadi-
sylate to the dihydrochloride in an excipient compatibility study. A
second batch of napadisylate was crystallized directly from the
dihydrochloride to produce material containing no detectable level
of amine 2. A mannitol:avicel mixture (1:1) with a 10% API loading
was prepared as an initial formulation blend. To this mixture, pH
modifiers were added. Citric acid was used as an acidifier, while
magnesium stearate was added to increase the alkalinity. The pH of
each blend was measured by preparing aqueous slurries according
Percent area/area of primary amine (2), determined by HPLC.
NT, not detected; RH, relative humidity.
a
Change from initial in parenthesis.
encouraging, but as our primary concern was around the stability of
the drug product, we were eager to see if the napadisylate would
show improved stability in the presence of excipients as well.
2
0
Excipient Compatibility of the Napadisylate
to the method described by Merritt et al.
The prepared blends were then stressed in the solid state for 4
weeks and analyzed for impurities by HPLC. The results are shown
in Table 5.
The data shown in Table 5 clearly demonstrate the impact of pH
modifiers. The dihydrochloride was more stable in the blends
It is often the case that apparently stable drug substances
become unstable as drug products once they are in physical contact
with excipients.24-28 This proved true in the present case, as
excipient compatibility studies conducted with the dihydrochloride
Table 5
Excipient Compatibility Results
Blend
Blend pH
Storage Condition (4 wk)
Amine 2 (% Area/Area)
Dihydrochloride
Napadisylate
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
Mannitol:avicel (1:1)
5.0
2.1
7.5
30 C, 55% RH, open vial
4
1.17
0.81
ND
<0.05
6.07
2.21
ND
ND
ND
ND
ND
ND
a
0 C, amb , closed vial
Mannitol:avicel:citric acid (1:1:0.1)
Mannitol:avicel:magnesium stearate (1:1:0.04)
30 C, 55% RH, open vial
4
0 C, amb, closed vial
30 C, 55% RH, open vial
4
0 C, amb, closed vial
NT, not detected; RH, relative humidity.
a
Ambient humidity.

J.M. Campbell et al. / Journal of Pharmaceutical Sciences xxx (2019) 1-7
7
modified with citric acid and less stable in the blends modified with
magnesium stearate. Thus, control of the microenvironmental pH
of the drug product provides an opportunity to improve the
chemical stability of 1.
2. Greenspan L. Humidity fixed points of binary saturated aqueous solutions. J Res
Natl Bur Stand Sec A Phys Chem. 1977;81A(1):89-96.
3. de Meijere A. Bonding properties of cyclopropane and their chemical conse-
quences. Angew Chem Int Ed Engl. 1979;18(11):809-826.
4. Reissig HU, Zimmer R. Donor-acceptor-substituted cyclopropane derivatives
and their application in organic synthesis. Chem Rev. 2003;103(4):1151-1196.
With respect to the napadisylate, the salt was more stable than
the dihydrochloride in every blend evaluated and showed no
degradation during the study. Evidently, the napadisylate is either
not susceptible to changes in formulation pH or was not stressed
harshly enough during the excipient compatibility study. None-
theless, the reduced solubility of the napadisylate clearly improved
the chemical stability of the API in the presence of excipients.
In the end, a drug product tablet was developed using the
napadisylate salt of 1 as well as acidification of the formulation
blend. The resulting tablet was found to have sufficient chemical
stability to permit long-term storage at room temperature.
While we think that the results of our investigation plainly
highlight the potential impact of microenvironmental pH on the
chemical stability of the drug product, it should be noted that hy-
drolytic degradation mechanisms may not have a straightforward
relationship with pH in the solid state. Other factors could
contribute depending strongly on the properties of the molecule
5
. Reißig H-U. Donor-acceptor-substituted cyclopropanes: versatile building
blocks in organic synthesis. In: Small Ring Compounds in Organic Synthesis III.
Topics in Current Chemistry. 144. Berlin, Heidelberg: Springer; 1988:73-135.
. Schneider TF, Kaschel J, Werz DB. A new golden age for donor-acceptor cy-
clopropanes. Angew Chem Int Ed Engl. 2014;53(22):5504-5523.
. Danishefsky S, Dynak J, Yamamoto M. Fluxional isomerism of a malonate anion
- an instance of intramolecular homoconjugate addition. J Chem Soc Chem
Commun. 1973;3:81-82.
. Maslivetc VA, Rubina M, Rubin M. One-pot synthesis of GABA amides via the
nucleophilic addition of amines to 3,3-disubstituted cyclopropenes. Org Biomol
Chem. 2015;13(34):8993-8995.
. Ozuduru G, Schubach T, Boysen MM. Enantioselective cyclopropanation of in-
doles: construction of all-carbon quaternary stereocenters. Org Lett.
6
7
8
9
2012;14(19):4990-4993.
10. de Nanteuil F, De Simone F, Frei R, Benfatti F, Serrano E, Waser J. Cyclization
and annulation reactions of nitrogen-substituted cyclopropanes and cyclo-
butanes. Chem Commun (Camb). 2014;50(75):10912-10928.
1
1. Rassadin VA, Six Y. Ring-opening, cycloaddition and rearrangement reactions
of nitrogen-substituted cyclopropane derivatives. Tetrahedron. 2016;72(32):
4701-4757.
1
1
1
1
2. Banning JE, Gentillon J, Ryabchuk PG, et al. Formal substitution of bromocy-
clopropanes with nitrogen nucleophiles. J Org Chem. 2013;78(15):7601-7616.
3. Kuehne ME, King JC. Cyclopropylamines as intermediates in a new method for
alkylation of aldehydes and ketones. J Org Chem. 1973;38(2):304-311.
4. Walborsky HM, Ronman PE. Cyclopropyl amine rearrangement. J Org Chem.
2
7,31
and the excipients present.
Conclusions
1973;38(24):4213-4214.
5. Tada T, Ishida Y, Saigo K. Ring-opening reaction of cyclopropanated [60]ful-
lerenes: unexpected transformation of methano[60]fullerenes having an
electron-donating group on the methano-bridge carbon. Org Lett. 2007;9(11):
The cyclopropyl amine moiety in GSK2879552 (1) degrades
hydrolytically in high pH conditions to give a primary amine (2) and
hydrocinnamaldehyde (3). The chemical stability of the molecule
impacted the shelf life of the drug product, and it was demon-
strated that this could be improved in 2 ways. First, selecting a salt
with reduced aqueous solubility, low saturated aqueous pH, and
minimal hygroscopicity improved the stability of the API in the
solid drug product even in the presence of excipients. Second, we
showed that acidification of the drug product provides an addi-
tional means of improving the chemical stability of 1.
These changes were made in a rational manner based on a
thorough knowledge of the solution-phase degradation chemistry,
and the results highlight the benefits of understanding degradation
mechanisms in drug development.
As drug product development continued, other factors such as
API particle characteristics, mechanical stress, API loading, and
2083-2086.
1
1
1
6. Connors KA, Amidon GL, Stella VJ. Chemical Stability of Pharmaceuticals: A
Handbook for Pharmacists. 2 ed. New York: John Wiley & Sons; 1986:43-54.
7. Guerrieri P, Taylor LS. Role of salt and excipient properties on disproportion-
ation in the solid-state. Pharm Res. 2009;26(8):2015-2026.
8. Stephenson GA, Aburub A, Woods TA. Physical stability of salts of weak bases in
the solid-state. J Pharm Sci. 2011;100(5):1607-1617.
19. Thakral NK, Kelly RC. Salt disproportionation: a material science perspective.
Int J Pharm. 2017;520(1-2):228-240.
2
2
0. Merritt JM, Viswanath SK, Stephenson GA. Implementing quality by design in
pharmaceutical salt selection: a modeling approach to understanding dispro-
portionation. Pharm Res. 2013;30(1):203-217.
1. Thakral NK, Behme RJ, Aburub A, et al. Salt disproportionation in the solid
state: role of solubility and counterion volatility. Mol Pharm. 2016;13(12):
4141-4151.
22. Guerrieri P, Jarring K, Taylor LS. Impact of counterion on the chemical stability
of crystalline salts of procaine. J Pharm Sci. 2010;99(9):3719-3730.
2
3. Guerrieri P, Rumondor AC, Li T, Taylor LS. Analysis of relationships between
solid-state properties, counterion, and developability of pharmaceutical salts.
AAPS PharmSciTech. 2010;11(3):1212-1222.
packaging were evaluated for their potential to impact the long-
term stability of 1.2
4,28
24. Antipas AS, Landis MS, Wuelfing WP. Solid-state excipient compatibiltiy
testing. In: Baertschi SW, Alsante KM, Reed RA, eds. Pharmaceutical Stress
Testing: Predicting Drug Degradation. 2nd ed. New York: Informa Healthcare;
2011:286-321.
Ultimately, a tablet formulated with acidi-
fied excipients and the napadisylate salt of 1 was developed, which
exhibited sufficient chemical stability to permit long-term storage
at room temperature.
25. Hotha KK, Roychowdhury S, Subramanian V. Drug-excipient interactions: case
studies and overview of drug degradation pathways. Am
016;07(01):107-140.
J Anal Chem.
2
Acknowledgments
2
6. Bharate SS, Bharate SB, Bajaj AN. Incompatibilities of pharmaceutical excipients
with active pharmaceutical ingredients: a comprehensive review. J Excip Food
Chem. 2010;1(3):3-26.
The authors acknowledge GlaxoSmithKline for its financial
support of this work. They also acknowledge Iain Reid, Stacy No-
bles, Stephanie Shore, Zhequiong Wu, Rachel Graves, Rennan Pan,
and Ross Woods for their technical assistance and helpful
discussions.
2
2
2
3
7. Narang AS, Desai D, Badawy S. Impact of excipient interactions on solid dosage
form stability. Pharm Res. 2012;29(10):2660-2683.
8. Gressl C, Brunsteiner M, Davis A, et al. Drug-excipient interactions in the solid
state: the role of different stress factors. Mol Pharm. 2017;14(12):4560-4571.
9. Badawy SI, Hussain MA. Microenvironmental pH modulation in solid dosage
forms. J Pharm Sci. 2007;96(5):948-959.
0. Zannou EA, Ji Q, Joshi YM, Serajuddin AT. Stabilization of the maleate salt of a
basic drug by adjustment of microenvironmental pH in solid dosage form. Int J
Pharm. 2007;337(1-2):210-218.
References
1
. Mohammad HP, Smitheman KN, Kamat CD, et al. A DNA hypomethylation
signature predicts antitumor activity of LSD1 inhibitors in SCLC. Cancer Cell.
31. Kornblum SS, Zoglio MA. Pharmaceutical heterogeneous systems I Hydrolysis
of aspirin in combination with tablet lubricants in an aqueous suspension.
J Pharm Sci. 1967;56(12):1569-1575.
2015;28(1):57-69.