Efficient synthesis of ESI-09, a novel non-cyclic nucleotide EPAC antagonist

Article abstract of DOI:10.1016/j.tetlet.2013.01.024

A concise and efficient synthetic approach to producing a novel non-cyclic nucleotide EPAC antagonist ESI-09 and its new analogs is reported. Key features of the synthesis include a mild and reliable one-pot procedure for an isoxazole synthon, as well as a modified one-pot protocol for the cyanomethyl ketone key intermediate. The synthesis requires inexpensive starting materials and only three linear steps for the completion in a total yield of 53%.

Full text of DOI:10.1016/j.tetlet.2013.01.024

Tetrahedron Letters 54 (2013) 1546–1549
Contents lists available at SciVerse ScienceDirect
Tetrahedron Letters
journal homepage: www.elsevier.com/locate/tetlet
Efficient synthesis of ESI-09, a novel non-cyclic nucleotide EPAC antagonist
Haijun Chen a, Chunyong Ding a, Christopher Wild a, Huiling Liu a, Tianzhi Wang b, Mark A. White b,
Xiaodong Cheng a,b, Jia Zhou a,
⇑
a Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, TX 77555, United States
b Sealy Center for Structural Biology and Molecular Biophysics, University of Texas Medical Branch, Galveston, TX 77555, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
A concise and efficient synthetic approach to producing a novel non-cyclic nucleotide EPAC antagonist
ESI-09 and its new analogs is reported. Key features of the synthesis include a mild and reliable one-
pot procedure for an isoxazole synthon, as well as a modified one-pot protocol for the cyanomethyl
ketone key intermediate. The synthesis requires inexpensive starting materials and only three linear
steps for the completion in a total yield of 53%.
Received 3 December 2012
Revised 4 January 2013
Accepted 8 January 2013
Available online 21 January 2013
Ó 2013 Elsevier Ltd. All rights reserved.
Keywords:
EPAC antagonist
ESI-09
Isoxazole
Modified Kowalski’s protocol
Cyanomethyl ketone
Pancreatic ductal adenocarcinoma (PDA) is one of the leading
causes of cancer death.1 Despite extensive research over the years
in terms of therapeutic strategies, no effective treatment has come
forward for this universally lethal disease.2 It is urgently needed to
develop novel chemical probes and effective therapeutics that are
based on new insights with better understanding of the molecular
mechanism of pancreatic cancer. The exchange protein directly
activated by cAMP (EPAC),3,4 especially EPAC1, is over-expressed
in human PDA specimens, but the functional roles of this over-
expression have never been investigated.5 Our team has previously
reported the identification and characterization of a high through-
put screening hit ESI-09 (1, chemical structure shown in Fig. 1), a
non-cyclic nucleotide small molecule that specifically inhibits
EPAC1 and EPAC2.6,7 Using ESI-09 as a small molecular chemical
probe, we demonstrated that EPAC1 plays an important role in
pancreatic cancer cell migration and invasion, suggesting EPAC1
may represent an attractive target for novel therapeutic strategies
for treating PDA.7
O
NC
HN
N
O
N
Cl
ESI-09 (1)
Figure 1. Structure of ESI-09 (1).
Our retrosynthetic strategy is outlined in Scheme 1. Coupling of
the key intermediate 8 with 3-chlorobenzenediazonium chloride
10 was expected to afford the desired product 1.8,9 Cyanomethyl
derivative 8 could be formed from ethyl ester 6.10 Isoxazole 6
would be derived from the commercially available pinacolone 2
and oxalic acid diethyl ester 3.11
We first prepared the 3,5-disubstituted isoxazole 6 in two steps
(Scheme 2). Starting from the commercially available and inexpen-
sive pinacolone 2, oxalic acid diethyl ester 3 and NaH (60%), b-dike-
tone 4 was readily prepared in 75% yield after purification. Further
treatment of b-diketone 4 with hydroxylamine hydrochloride for
about 2 h at room temperature afforded the intermediate oxime
5. Upon stirring at 75 °C for another 4 h, followed by simple extrac-
tion procedures and short column silica gel chromatography, the
desired product 6 was obtained in 90% yield. Despite the good
yields for preparation of isoxazole 6, this two-step procedure
required two column purification steps. Based on the reagents
In our ongoing research to develop small molecules with high
potency, specificity, and druglike properties based upon ESI-09, it
is imperative to establish a practical and efficient method to read-
ily access 1 suitable for a large scale as a reference compound as
well as to further evaluate its pharmacological properties in vivo.
Herein, we describe our optimized synthetic approaches to 1 start-
ing from pinacolone in high overall yields.
⇑
Corresponding author. Tel.: +1 409 772 9748; fax: +1 409 772 9818.
E-mail address: jizhou@utmb.edu (J. Zhou).
0040-4039/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tetlet.2013.01.024

H. Chen et al. / Tetrahedron Letters 54 (2013) 1546–1549
1547
Table 1
O
Optimization of reaction conditions of bromomethyl ketone 7a
Cl
NC
HN
O
N
Products (yieldb)
NC
O
Entry
Base (equiv)
BrCH2Br (equiv)
Time (h)
N
N
+
O
1
2
3
4
5
6
7
LDA (2.0)
2.0
2.0
2.0
2.0
2.0
2.0
1.2
1
1
1
2
2
2
2
7 (6%), 11 (trace)
7 (11%), 11 (trace)
7 (25%), 11 (25%)
7 (57%), 11 (5%)
7 (65%), 11 (3%)
7 (73%), 11 (trace)
7 (84%), 11 (trace)
N
N
n-BuLi (2.0)
MeLi (2.0)
MeLi (2.5)
MeLi (3.0)
MeLi (4.0)
MeLi (3.0)
Cl
Cl
10
1
8
a
O
The concentration was 0.25 M in THF.
Isolated yield.
b
O
N
EtO
OEt
O
+
EtO
O
O
and the desired product 7 was obtained in 73% yield by using
4.0 equiv of MeLi (entry 6). Considering the mechanism of this
reaction,12 we reasoned that 1.2 equiv of dibromomethane would
be enough for this reaction. As expected, the desired product 7
was obtained in 84% yield under an optimized condition by using
1.2 equiv of dibromomethane and 3.0 equiv of MeLi (entry 7).
With bromomethyl ketone 7 in hand, we used trimethylsilyl
cyanide (TMSCN) as cyanide source to covert 7 into cyanomethyl
derivative 8.13 In the presence of tetrabutylammonium fluoride
(TBAF), cyanomethyl ketone 8 was smoothly generated in the sol-
vent of CH3CN at room temperature for about 3 h (Scheme 4). This
reaction was clear with only a small polar spot observed via TLC.
However, it was noteworthy that cyanomethyl ketone 8 could
decompose upon purification by standard column chromatography
(unstable in silica gel column or neutral Al2O3 column) and there-
fore, we used this intermediate directly for the next step without
further purification.
2
3
6
Scheme 1. Retrosynthetic analysis of ESI-09 (1).
O
O
NaH (60%)
OEt
+
EtO
OEt
THF
O
2
O
O
OH
3
4
EtO2C
O
N
NH2OH.HCl
reflux
O
OEt
OH
EtOH/THF, rt
O
N
5
6
Scheme 2. One-pot synthesis of 5-tert-butylisoxazole-3-carboxylic acid ethyl ester
6.
Aryldiazonium salt 10 was prepared from 3-chloroaniline 9 by
using sodium nitrite in the presence of hydrochloric acid at 0 °C
for 2 min. Direct coupling of the crude cyanomethyl ketone 8 with
aryldiazonium chloride 10 afforded the desired product 1 in 51%
yield in two steps (Scheme 5). The chemical structure of 1 was fur-
ther confirmed by single-crystal X-ray structural analysis.14
Despite the synthesis of ESI-09 in moderate yields (an overall
yield of 37% in four steps), we attempted to shorten the synthetic
route by means of exploring one-pot procedure for converting
isoxazole 6 into cyanomethyl ketone 8 according to our modified
Kowalski’s protocol for bromomethyl ketone (Table 1, entry 7).
Replacing CH3CN with dibromomethane, we found that 6 was con-
verted into 8 in high yield by using 2.0 equiv MeLi as the base
(Scheme 6).15 Consistent with experimental observations from
the synthesis of bromomethyl ketone 7, it was noted that MeLi is
the optimal base for preparation of ketonitrile 8 (less than 50%
yields by using n-BuLi or LDA as the base under the same reaction
condition). Using the same subsequent coupling reaction, desired
product 116 was obtained in 61% yield in two steps from isoxazole
6. To further explore the synthetic potential for generating ESI-09
analogs using this optimized approach, a series of new derivatives
12a–d (Scheme 7) have been synthesized in good yields (62–76%,
two steps).
which were used in this procedure, we reasoned that we could use
hydroxylamine hydrochloride directly without further purification
of b-diketone 4. In this event, treatment of the reaction solution of
b-diketone 4 with the solution of hydroxylamine hydrochloride in
ethanol afforded the desired product 6 after refluxing at 75 °C for
24 h. To our delight, we found that only a single clear spot was
detectable by TLC. After short column silica gel chromatography,
isoxazole 6 was obtained in 87% yield.
Next, we followed Kowalski’s protocol12 to synthesize the key
intermediate 7, which was carried out at À78 °C in the presence
of 2.0 equiv of LDA and 2.0 equiv of dibromomethane (Scheme 3).
Based upon our findings, the reaction was incomplete after 1 h,
and the product was a complex reaction mixture of several compo-
nents. After purification with silica gel column, the desired bromo-
methyl ketone 7 was only obtained in 6% yield and we also isolated
the trace amount of the dibromo by-product 11 (Table 1, entry 1).
We then investigated the different bases for this reaction and the
results showed that MeLi was the best base (entry 3), but the yield
was still low (only 25% yield). It was noteworthy that this modified
condition (entry 3) resulted in a clear reaction while approximately
50% of the starting material was recovered and the by-product 11
was also isolated in 25% yield. To improve the yield of this reaction,
we increased the amount of MeLi and extended the reaction time
from 1 to 2 h (entry 4–6). To our delight, the reaction was also clear
In conclusion, we have developed a straightforward method to
synthesize 1 starting from pinacolone in an excellent overall yield
of 53% in three steps. The efficient approach included an one-pot
O
O
O
O
Br
Br
EtO2C
NC
Br
N
N
N
BrCH2Br
N
N
TMSCN
+
O
O
O
O
Br
O
Base, THF
TBAF, CH3CN
11
6
7
7
8
Scheme 3. Synthesis of 2-bromo-1-(5-tert-butylisoxazol-3-yl)ethanone 7.
Scheme 4. Synthesis of 3-(5-tert-butylisoxazol-3-yl)-3-oxo-propionitrile 8.

1548
H. Chen et al. / Tetrahedron Letters 54 (2013) 1546–1549
Supplementary data
Cl
N
Cl
NH2
Cl
N
1 N HCl
8, EtOH,
NaOAc
Supplementary data associated with this article can be found, in
NaNO2, H2O
the
online
version,
at
http://dx.doi.org/10.1016/
9
10
j.tetlet.2013.01.024.
O
References and notes
NC
N
O
1. Hariharan, D.; Saied, A.; Kocher, H. M. HPB (Oxford) 2008, 10, 58.
2. (a) Hezel, A. F.; Kimmelman, A. C.; Stanger, B. Z.; Bardeesy, N.; Depinho, R. A.
Genes Dev. 2006, 20, 1218; (b) Hidalgo, M. N. Engl. J. Med. 2010, 362, 1605.
3. (a) de Rooij, J.; Zwartkruis, F. J.; Verheijen, M. H.; Cool, R. H.; Nijman, S. M.;
Wittinghofer, A.; Bos, J. L. Nature 1998, 396, 474; (b) Kawasaki, H.; Springett, G.
M.; Mochizuki, N.; Toki, S.; Nakaya, M.; Matsuda, M.; Housman, D. E.; Graybiel,
A. M. Science 1998, 282, 2275.
N
HN
Cl
1
4. (a) Holz, G. G.; Chepurny, O. G.; Schwede, F. Cell. Signal. 2008, 20, 10; (b)
Gloerich, M.; Bos, J. L. Annu. Rev. Pharmacol. Toxicol. 2010, 50, 355; (c) Grandoch,
M.; Roscioni, S. S.; Schmidt, M. Br. J. Pharmacol. 2010, 159, 265; (d) Breckler, M.;
Berthouze, M.; Laurent, A. C.; Crozatier, B.; Morel, E.; Lezoualc’h, F. Cell. Signal.
2011, 23, 1257.
Scheme 5. Synthesis of 3-(5-tert-butylisoxazol-3-yl)-2-[(3-chlorophenyl)-hydraz-
ono]-3-oxo-propionitrile 1.
5. Lorenz, R.; Aleksic, T.; Wagner, M.; Adler, G.; Weber, C. K. Pancreas 2008, 37,
102.
O
NC
EtO2C
N
N
6. (a) Tsalkova, T.; Mei, F. C.; Cheng, X. PLoS ONE 2012, 7, e30441; (b) Tsalkova, T.;
Mei, F. C.; Li, S.; Chepurny, O. G.; Leech, C. A.; Liu, T.; Holz, G. G.; Woods, V. L.,
Jr.; Cheng, X. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 18613; (c) Chen, H.; Tsalkova,
T.; Mei, F. C.; Hu, Y.; Cheng, X.; Zhou, J. Bioorg. Med. Chem. Lett. 2012, 22, 4038.
7. Almahariq, M.; Tsalkova, T.; Mei, F. C.; Chen, H.; Zhou, J.; Sastry, S. K.; Schwede,
F.; Cheng, X. Mol. Pharmacol. 2013, 83, 122.
8. (a) Abdel-Motaleb, R. M.; Makhloof, A. A.; Ibrahim, H. M.; Elnagdi, M. H. J.
Heterocycl. Chem. 2006, 43, 931; (b) Aziz, S. I.; Anwar, H. F.; Fleita, D. H.; Elnagdi,
M. H. J. Heterocycl. Chem. 2007, 44, 725.
CH3CN
O
O
MeLi, THF
Yield up to 90%
6
8
Scheme 6. Synthesis of 3-(5-tert-butylisoxazol-3-yl)-3-oxo-propionitrile 8 from 6.
9. (a) Hassaneen, H. M. E. Synth. Commun. 2007, 37, 3579; (b) Behbehani, H.;
Ibrahim, H. M.; Makhseed, S.; Mahmoud, H. Eur. J. Med. Chem. 2011, 46, 1813.
10. (a) Arseniyadis, S.; Kyler, K. S.; Watt, D. S. Org. React. (N.Y.) 1984, 31, 1; (b)
Katritzky, A. R.; Abdel-Fattah, A. A. A.; Wang, M. J. Org. Chem. 2003, 68, 4932; (c)
Ji, Y.; Trenkle, W. C.; Vowles, J. V. Org. Lett. 2006, 8, 1161.
11. (a) PinhoeMelo, T. M. V. D. Curr. Org. Chem. 2005, 9, 925; (b) Lepage, F.;
Tombret, F.; Cuvier, G.; Marivain, A.; Gillardin, J. M. Eur. J. Med. Chem. 1992, 27,
581.
O
NC
N
O
N
HN
NH2
a) 1 N HCl, NaNO2, H2O
R
b) 8, NaOAc, EtOH
12. Kowalski, C. J.; Haque, M. S. J. Org. Chem. 1985, 50, 5140.
13. Soli, E. D.; Manoso, A. S.; Patterson, M. C.; DeShong, P.; Favor, D. A.;
Hirschmann, R.; Smith, A. B., III J. Org. Chem. 1999, 64, 3171.
R
62-76% yield
(two steps from 6)
12a: R = 2-Cl, 76%
14. The crystallographic data have been deposited in full at the Cambridge
Crystallographic Data Centre, reference CCDC 913321, accessible at http://
www.ccdc.cam.ac.uk.
12b: R = 4-Cl, 75%
12c: R = 2,5-diCl, 62%
12d: R = H, 76%
15. 1H and 13C NMR spectra (see spectra of crude product in the Supplementary
data) showed that the crude residue was only contaminated with some solvent
impurities and suitable for direct use in the next step.
Scheme 7. Synthesis of ESI-09 analogs 12a–d.
16. One-pot synthesis of 5-tert-butylisoxazole-3-carboxylic acid ethyl ester (6). To
a solution of pinacolone (5.0 g, 50.0 mmol) in THF (100 mL) was added NaH
(60%) (2.2 g, 55.0 mmol) at 0 °C. The reaction mixture was stirred at rt for about
30 min. Diethyl oxalate (7.3 g, 50.0 mmol) was added at 0 °C and the mixture
was then stirred at rt overnight. To the resulting solution, hydroxylamine
hydrochloride (3.8 g, 55.0 mmol) in ethanol (100 mL) was added dropwise. The
mixture was heated at reflux for 16 h. After this time the sodium chloride was
removed by filtration and the filtrate was concentrated under reduced
pressure. The residue was purified by short column chromatography on silica
gel eluting with hexane/ethyl acetate (4:1) to provide the title compound as a
colorless oil (8.57 g, 87%). 1H NMR (600 MHz, CDCl3) d 6.37 (s, 1H), 4.43 (q, 2H,
J = 7.2 Hz), 1.41 (t, 3H, J = 7.2 Hz), 1.37 (s, 9H). 13C NMR (150 MHz, CDCl3) d
183.3, 160.5, 156.2, 99.3, 62.1, 33.1, 28.9, 14.3. HRMS (ESI) calcd for C10H16NO3
198.1125 (M+H)+, found 198.1131.
procedure for the isoxazole synthon, followed by a modified proto-
col for the key cyanomethyl ketone and the subsequent coupling
step. The described synthetic strategies and optimized reaction
conditions may have general applications in the convenient prepa-
ration of various isoxazole and cyanomethyl ketone building
blocks. Moreover, the established concise and efficient synthesis
of the novel EPAC antagonist ESI-09 will facilitate its further
in vitro and in vivo pharmacological evaluations, as well as the
ongoing synthesis of new analogs for the structure–activity rela-
tionship study.
2-Bromo-1-(5-tert-butylisoxazol-3-yl)ethanone (7). To
a solution of 5-tert-
butylisoxazole-3-carboxylic acid ethyl ester (0.5 g, 2.5 mmol) in anhydrous
tetrahydrofuran (10 mL) was added dibromomethane (0.52 g, 3.0 mmol) under
nitrogen. The mixture was cooled to À78 °C and 1.6 M methyllithium in diethyl
ether (4.7 mL, 7.5 mmol) was added dropwise. The solution was stirred at
À78 °C for 40 min and then quenched with acetic acid (0.45 g, 7.5 mmol). The
mixture was warmed to 0 °C and poured onto ice/water (20 mL) and extracted
with ethyl acetate (50 mL). The organic layer was dried over Na2SO4, filtered,
and concentrated under reduced pressure. The residue was purified by short
column chromatography on silica gel eluting with hexane/ethyl acetate (4:1)
to provide the title compound as a colorless oil (524 mg, 84%). 1H NMR
(600 MHz, CDCl3) d 6.40 (s, 1H), 4.57 (s, 2H), 1.38 (s, 9H). 13C NMR (150 MHz,
CDCl3) d 185.7, 183.9, 159.7, 97.7, 33.2, 31.7, 28.9. HRMS (ESI) calcd for
C9H13BrNO2 246.0124 (M+H)+, found 246.0131.
Notes
The authors declare no competing financial interest.
Acknowledgments
This work was supported by Grants P30DA028821,
R21MH093844 (J.Z.), and R01GM066170, R21NS066510 (X.C.) from
the National Institute of Health, R. A. Welch Foundation Chemistry
and Biology Collaborative Grant from Gulf Coast Consortia (GCC)
for Chemical Genomics, Sealy and Smith Foundation grant (to the
Sealy Center for Structural Biology and Molecular Biophysics), John
Sealy Memorial Endowment Fund, and the Center for Addiction
Research (CAR) at the University of Texas Medical Branch.
2,2-Dibromo-1-(5-tert-butylisoxazol-3-yl)ethanone (11). Colorless oil. 1H NMR
(600 MHz, CDCl3) d 6.40 (s, 1H), 4.90 (s, 1H), 1.38 (s, 9H). 13C NMR (150 MHz,
CDCl3) d 193.8, 183.6, 159.2, 97.2, 66.7, 33.1, 28.9. HRMS (ESI) calcd for
C9H12Br2NO2 323.9229 (M+H)+, found 323.9320.
3-(5-tert-Butylisoxazol-3-yl)-3-oxo-propionitrile (8). To a solution of 2-bromo-
1-(5-tert-butylisoxazol-3-yl)ethanone (0.62 g, 2.5 mmol) in 10 mL of
acetonitrile was added trimethylsilyl cyanide (0.25 g, 2.5 mmol), followed by
2.5 mL (2.5 mmol) of tetrabutylammonium fluoride solution (1 M in THF). The

H. Chen et al. / Tetrahedron Letters 54 (2013) 1546–1549
1549
reaction mixture was stirred at rt for 0.5 h, and then diluted with brine
(10 mL). The product was extracted with ethyl acetate (50 mL). The organic
layer was dried over Na2SO4, filtered, and concentrated under reduced
pressure. The residue was obtained as a yellow oil and used directly for next
step without further purification. It is noteworthy that cyanomethyl ketone 8
was unstable in silica gel column or neutral Al2O3 column. The pure compound
for characterization was purified by preparative plate chromatography (silica
gel, hexane/ethyl acetate 2:1) to obtain 8 as a yellow oil. 1H NMR (600 MHz,
CDCl3) d 6.42 (s, 1H), 4.18 (s, 2H), 1.38 (s, 9H). 13C NMR (150 MHz, CDCl3) d
184.6, 182.0, 160.0, 112.9, 97.3, 33.2, 29.9, 28.8. HRMS (ESI) calcd for
reduced pressure. The crude residue (490 mg) was obtained as a yellow oil and
directly used for next step without further purification.
3-(5-tert-Butylisoxazol-3-yl)-2-[(3-chlorophenyl)-hydrazono]-3-oxo-
propionitrile (1). To a solution of 3-chloroaniline (30 mg, 0.24 mmol) in H2O
(1 mL cooled to À5 °C) was added 0.24 mL of 1 N HCl (aq). To the resulting
acidic aniline solution, 1 mL solution of sodium nitrite (16 mg, 0.24 mmol) in
H2O was added dropwise to generate the aryldiazonium salt solution. To the
aryldiazonium salt solution was added sodium acetate (33 mg, 0.4 mmol),
followed by 1 mL solution of crude 3-(5-tert-butylisoxazol-3-yl)-3-oxo-
propionitrile (38 mg, 0.2 mmol) in ethanol. The reaction mixture was stirred
at 0 °C for 5 min, and then poured onto H2O (2 mL) and extracted with ethyl
acetate (20 mL). The organic layer was dried over Na2SO4, filtered, and
concentrated under reduced pressure. The residue was purified by short
column chromatography on silica gel eluting with hexane/ethyl acetate (2:1)
to provide the desired product ESI-09 (40 mg, 61%, two steps from 6) as a
yellow solid (mp 146–147 °C). HPLC purity 99.6% (tR = 21.72 min). 1H NMR
(600 MHz, DMSO-d6) d 12.70 (br s, 1H), 7.44–7.47 (m, 3H), 7.25–7.26 (m, 1H),
6.70 (s, 1H), 1.39 (s, 9H). 13C NMR (150 MHz, DMSO-d6) d 181.1, 179.4, 160.1,
143.6, 134.0, 131.2, 125.1, 116.2, 115.8, 113.4, 110.5, 100.4, 32.5, 28.5. HRMS
(ESI) calcd for C16H16ClN4O2 331.0956 (M+H)+, found 331.0969.
C
10H13N2O2 193.0972 (M+H)+, found 193.1020.
One-pot synthesis of 3-(5-tert-butylisoxazol-3-yl)-3-oxo-propionitrile (8) from
6. To a solution of CH3CN (0.41 g, 10.0 mmol) in anhydrous THF (5 mL) was
added 1.6 M methyllithium in diethyl ether (3.1 mL, 5.0 mmol) at À78 °C under
nitrogen. The mixture was stirred at À78 °C for 0.5 h, and 5-tert-
butylisoxazole-3-carboxylic acid ethyl ester (0.5 g, 2.5 mmol) in THF (5 mL)
was then added dropwise. The solution was stirred at À78 °C for 1 h and then
quenched with acetic acid (0.3 g, 5.0 mmol). The mixture was warmed to 0 °C
and poured onto ice/water (10 mL) and extracted with ethyl acetate (50 mL).
The organic lay was dried over Na2SO4, filtered, and concentrated under