Chemistry of the 2,5-didehydropyridine biradical: Computational, kinetic, and trapping studies toward drug design

Article abstract of DOI:10.1021/ja9730223

A combined computation, kinetic, and trapping study of the 2,5-didehydropyridine biradical finds the hydrogen abstraction reaction to be unable by protonation. The observation of small amounts of pyridine products in the thermolysis of a C,N-dialkynylimin

Full text of DOI:10.1021/ja9730223

376
J. Am. Chem. Soc. 1998, 120, 376-385
Chemistry of the 2,5-Didehydropyridine Biradical: Computational,
Kinetic, and Trapping Studies toward Drug Design
Johannes Hoffner, Marc J. Schottelius, Derek Feichtinger, and Peter Chen*
Laboratorium fu¨r Organische Chemie, Eidgeno¨ssische Technische Hochschule (ETH), Zu¨rich, Switzerland
ReceiVed August 27, 1997
Abstract: A combined computation, kinetic, and trapping study of the 2,5-didehydropyridine biradical finds
the hydrogen abstraction reaction to be tunable by protonation. The observation of small amounts of pyridine
products in the thermolysis of a C,N-dialkynylimine, or azaenediyne, only when the reaction occurs in the
presence of moderate amounts of protic acid, is consistent with qualitative theoretical predictions as well as
ab initio calculations at the CASSCF and CASMP2 levels. The implication of these findings for a more
selective antitumor agent is discussed.
Introduction
achieve this end through immunoconjugates of the calicheamicin
chromophore to a monoclonal antibody9 have not been entirely
successful.
We report a combined computational, kinetic, and trapping
study of the reactions of the C,N-dialkynylimines,1 or azaene-
diynes, with an emphasis on the delineation of the conditions
under which the chemistry of an intermediate 1,4-arene biradical
might direct us toward a DNA-damaging agent with greater
selectivity against tumor cells as compared to p-benzyne. The
latter biradical occurs in the activated form of the enediyne
natural products2 such as calicheamicin, dynemicin, and espe-
ramicin and has been implicated as the key structural element
conferring cytotoxicity to the compounds in Scheme 1.3
In the present study, the desired end is approached differently
through a modification to the enediyne “warhead” common to
all the natural products so as to generate a modified biradical
intermediate whose DNA-damaging ability would be enhanced
for tumor cells.10 The operational model is a qualitative
quantum chemical picture of reactivity patterns in singlet
carbenes11 and biradicals, on which basis we had conducted
kinetic studies in solution on the hydrogen abstraction rates for
substituted p-benzyne biradicals.12 The prediction of slow
hydrogen abstraction by singlet biradicals was experimentally
confirmed in that work, leading us to the present prediction that
biradicals based on a 2,5-didehydropyridine nucleus could show
pH-dependent hydrogen abstraction ability (Scheme 2).
The present computational, kinetic, and trapping study, and
accompanying synthetic project, seeks to verify the prediction
that the 2,5-didehydropyridine biradical does abstract hydrogens
more readily at the low pH values and, hence, can serve as the
prototype for a DNA-damaging agent more selective for tumor
cells.
These natural products, as well as their more recently
discovered cousins, e.g. kedarcidin,4 C-1027,5 and namenami-
cin,6 were promising candidates as chemotherapeutic agents.
The compounds displayed high activity against a number of
tumor cell lines,7 both in vitro and in vivo, but unfortunately
also showed unacceptable toxic side effects in both animal and
human trials.8 Further development of this family of compounds
as drugs requires, accordingly, a way to engineer greater
selectivity for tumor cells into the compounds. Attempts to
Experimental Section
(1) The first synthesis for this class of compounds has recently been
reported. David, W. M.; Kerwin, S. M. J. Am. Chem. Soc. 1997, 119, 1464.
(2) An entire monograph is devoted to the various aspects of enediyne
chemistry, and serves as an excellent introduction: Enediyne Antibiotics
as Antitumor Agents; Borders, D. B., Doyle, T. W., Eds.; Marcel Dekker:
New York, 1995.
(3) Goldberg, I. H.; Kappen, L. S. In Enediyne Antibiotics as Antitumor
Agents; Borders, D. B., Doyle, T. W., Eds.; Marcel Dekker: New York,
1995; Chapter 16.
(4) Leet, J. E.; Schroeder, D. R.; Langley, D. R.; Colson, K. L.; Huang,
S.; Klohr, S. E.; Lee, M. S.; Golik, J.; Hofstead, S. J.; Doyle, T. W.; Matsen,
J. A. J. Am. Chem. Soc. 1993, 115, 8432.
(5) Minami, Y.; Yoshida, K.; Azuma, R.; Saeki, M.; Otani, T. Tetrahe-
dron Lett. 1993, 2633. Yoshida, K.; Minami, Y.; Azuma, R.; Saeki, M.;
Otani, T. Tetrahedron Lett. 1993, 2637.
(6) McDonald, L. A.; Capson, T. L.; Krishnamurthy, G.; Ding, W. D.;
Ellestad, G. A.; Bernan, V. S.; Maiese, W. M.; Lassota, P.; Discafani, C.;
Kramer, R. A.; Ireland, C. M. J. Am. Chem. Soc. 1996, 118, 10898.
(7) Durr, F. E.; Wallace, R. E.; Testa, R. T.; Kuck, N. A. In Enediyne
Antibiotics as Antitumor Agents; Borders, D. B., Doyle, T. W., Eds.; Marcel
Dekker: New York, 1995; Chapter 8.
(8) Casazza, A. M.; Kelley, S. L. In Enediyne Antibiotics as Antitumor
Agents; Borders, D. B., Doyle, T. W., Eds.; Marcel Dekker: New York,
1995; Chapter 14.
The synthesis of C,N-dialkynylimines, 1a and 1b, was done
according to a route which closely resembles the synthesis published
recently by David and Kerwin.1 Alkynylation of the tosyl ester of a
C-alkynyloxime by an alkynyl cuprate reagent proceeds without
Beckmann rearrangement, as reported by Wuerthwein and Weigman13
for alkynylations of similar oxime esters, most likely through a free
radical coupling mechanism. Kinetic runs on the tandem ring closure/
ring opening of the azaenediynes were performed for 1b. Dilute
samples of 1b in n-heptane (0.14 mM) were thermolyzed in a constant
temperature oil bath (temperature stability ( 0.2 °C), removed, and
(9) Hinman, L. M.; Hamann, P. R.; Beyer, C. F.; Wallace, R.; Upeslacis,
J.; Adair, J.; Baker, T.; Mountain, A. Proc. Am. Assoc. Cancer Res. 1993,
34, 479.
(10) Chen, P. Angew. Chem. 1996, 108, 1584.
(11) The earlier work is covered in a review: Blush, J. A.; Clauberg,
H.; Kohn, D. W.; Minsek, D. W.; Zhang, X.; Chen, P. Acc. Chem. Res
1992, 25, 385. Recently extended: Taton, T. A.; Chen, P. Angew. Chem.,
Int. Ed. Engl. 1996, 35, 1011.
(12) Schottelius, M. J.; Chen, P. J. Am. Chem. Soc. 1996, 118, 4896.
(13) Wuerthwein, E.-U.; Weigman, R. Angew. Chem. 1987, 99, 918.
S0002-7863(97)03022-9 CCC: $15.00 © 1998 American Chemical Society
Published on Web 01/06/1998

Chemistry of the 2,5-Didehydropyridine Biradical
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 377
Scheme 1
Methyl (4-tert-Butylphenyl)ethynyl Ketone.14 A 9.28 g portion
of 4-tert-butylphenylacetylene (58.7 mmol) was dissolved in 100 mL
of freshly distilled THF and cooled to -78 °C. Then, 36.5 mL of 1.6
N n-butyllithium in hexane (58.4 mmol) was added slowly. The
solution was stirred for 90 min, after which 7.3 mL of BF3‚Et2O (8.2
g, 58.4 mmol) was added dropwise. A cold solution (-78 °C) of 8.8
g of acetic anhydride in 10 mL of THF was injected in one portion.
The resulting mixture was stirred for 15 min. The reaction was
quenched with 200 mL of 1 N NaOH solution and extracted with
hexane. The resulting turbid organic layer was filtered through Celite,
dried over MgSO4, and isolated by vacuum distillation (0.05 Torr/88
°C). Yield: 5.43 g, 27.2 mmol, 47%. Mp: 33-35 °C. IR (CCl4):
2966 s, 2906 m, 2203 s, 1676 s, 1504 m, 1358 m, 1282 m, 1157 m,
1106 m, 1018 m. 1H NMR (200 MHz, CDCl3): 7.55-7.34 (m, 4H),
2.45 (s, 3H), 1.33 (s, 9H). 13C NMR (50 MHz, CDCl3): 184.2, 154.1,
132.6, 125.3, 116.4, 90.6, 87.7, 34.6, 32.3, 30.6. MS (EI): 200 [M+;
22], 185 [M+ - 15; 100], 170 [M+ - 30; 10], 157 [M+ - 43; 18], 115
[M+ - 85; 20]. HRMS (EI): 200, exact mass found 200.1201, calcd
for C14H16O, 200.120 115. Rf (hexane/ethyl acetate, 4/1): 0.35.
Methyl (4-tert-Butylphenyl)ethynyl Ketoxime. Methyl (4-tert-
butylphenyl)ethynyl ketone (5.43 g, 27.2 mmol), 2.10 g of NH2OH‚
HCl, and 5.00 g of sodium acetate were suspended in 50 mL of ethanol
and stirred vigorously. The reaction was followed by TLC and stopped
after 2 h, after the ketone had fully reacted. The mixture was extracted
with water and then hexane/ethyl acetate, 4/1, and the organic layer
was dried over Na2SO4. The crude material was separated by column
chromatography (5.8 cm, 520 g of silica, hexane/ethyl acetate, 6/1).
Note that the oxime with the OH syn to the acetylenic linkage rapidly
cyclizes; hence only the anti isomer is isolated. Yield: 2.09 g, 9.70
mmol, 36%. IR (KBr): 3300 b, 3969 s, 2222 w, 1603 w, 1503 m,
1024 m. 1H NMR (200 MHz, CDCl3): 8.44 (s, 1H), 7.50-7.38 (m,
4H), 2.16 (s, 3H), 1.32 (s, 9H). 13C NMR (75 MHz, CDCl3): 153.4,
139.8, 132.2, 125.7, 118.7, 100.9, 79.8, 35.0, 31.2, 20.6. MS (EI):
Scheme 2
assayed by HPLC (Merck LaChrome, LiChrospher Si-60, 250/4 5 µm
column, UV detector fixed at 337 nm). The rearrangement was clean
and first-order, yielding 24 points at five temperatures over a 42 °C
range. These were fitted to the Arrhenius equation, yielding a linear
fit (correlation coefficient: 0.996) from which the activation parameters
were extracted in the ordinary way. The crystalline solid 1b was used
for the Arrhenius plot, instead of 1a, an oily liquid, because of its greater
ease in purification and handling.
Trapping of the a putative biradical intermediate at room temperature
and ∼100 °C was done by thermolysis of dilute (∼2 mM) solutions of
1a in sealed vials containing diisopropyl ether (solvent and hydrogen
atom donor), 2-fluoropyridine (solvent and buffer), and HBF4 (protic
acid) in varying proportions. The diisopropyl ether and fluoropyridine,
typically in a 1:1 volume-to-volume ratio, serve as the reaction medium.
After each reaction, the solutions were subjected to a minimal workup
(only enough to protect the GC/MS), consisting of neutralization with
solid K2CO3 and filtration through a short silica plug, prior to GC/MS
analysis (Fisons MD800, DB11 capillary column). Authentic samples
of 2-methyl-4,5-diphenylpyridine (4a), 1-cyano-1,2-diphenylpent-1-en-
3-yne (3a), phenylacetonitrile (5), and 1-phenylbut-1-yn-3-one (6)
(which are, respectively, the products of hydrogen abstraction by an
intermediate biradical, the cyclization/ring-opening product, and the
two hydrolysis products of 1a) were prepared to validate the GC/MS
identification of each compound by both retention time and mass
spectrometric fragmentation pattern. The azaenediynes themselves,
when run through the GC/MS, are indistinguishable from the product
nitriles, presumably because they rearrange rapidly in the 230 °C GC
injector port. Deuterium abstraction experiments were conducted in
the same manner, but with diisopropyl ether-d14 as the solvent.
215 [M+; 35], 200 [M+ - 15; 100], 184 [M+ - 31; 20], 170 [M+
-
45; 20]. HRMS (EI): 200, exact mass found 200.1075, calcd for
C13H14NO 200.107 539. Rf (hexane/ethyl acetate, 6/1): 0.375.
Methyl (4-tert-Butylphenyl)ethynyl Ketoxime Tosylate. The
oxime (2.09 g, 9.70 mmol) was dissolved in an ice-cold mixture of
pyridine and dichloromethane, 2.12 g of tosyl chloride was added, and
the reaction was monitored by TLC. After 5 h the solvents were
removed, and the reaction mixture was consecutively extracted with
400 mL of ethyl acetate and washed with 100 mL of 0.2 M HCl, 100
mL of saturated NaHCO3 solution, and 100 mL of brine. The organic
layer was dried with MgSO4, and the solvent was removed. The
recrystallization from hexane/dichloromethane yielded 1.96 g (55%)
(14) Brown, H. C.; Racherla, U. S.; Singh, S. M. Tetrahedron Lett. 1984,
25, 2411.

378 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Hoffner et al.
of crystals. Mp: 79-82 °C. IR (CCl4): 2968 s, 2870 m, 2234 m,
2191 m, 1598 m, 1387 s, 1180 s, 1096 m, 1023 m. 1H NMR (200
MHz, CDCl3): 7.92 (d, J ) 7.88, 2H), 7.5-7.3 (m, 6H), 2.47 (s, 3H),
2.18 (s, 3H), 1.35 (s, 9H). 13C NMR (50 MHz, CDCl3): 154.2, 150.0,
145.0, 132.8, 132.4, 129.6, 128.9, 125.6, 117.5, 103.3, 78.9, 34.9, 30.9,
and heated to 240 °C for 30 min, followed by 300 °C for 10 min. The
mixture was allowed to cool to room temperature and extracted with
25 mL of ethanol. The solvent was removed, and the crude product
was dissolved in ether. A 20 mL portion of 1 N HCl in ether was
added to precipitate the hydrochlorides. The solution was decanted,
and the insoluble salts were dissolved in ethanol. After ether was added,
the solution became turbid. The mixture was cooled to -78 °C and
allowed to warm to room temperature overnight. The resulting crystals
were the pure hydrochloride of 4a. The hydrochloride was extracted
with 25 mL of 1 N NaOH and ether, the organic layer was dried over
Na2SO4, and the solvent was removed. The pyridine was a waxy, white
solid with no detectable melting point. IR (CCl4): 3060 m, 3027 m,
2961 m, 2926 m, 1947 w, 1877 w, 1807 w, 1754 w, 1590 s, 1475 s,
1438 m, 1381 m, 1262 s. 1H NMR (300 MHz, CDCl3): 8.51 (s, 1H),
7.28-7.22 (m, 6H), 7.21 (s, 1H), 7.16-7.10 (m, 4H), 2.63 (s, 3H).
13C NMR (75 MHz, CDCl3): 157.6, 150.5, 148.4, 139.1, 138.0, 133.4,
130.1, 129.5, 128.4, 128.0, 127.3, 124.4, 111.2, 23.3. MS (EI): 245
[M+; 100], 244 [M+ -H; 65], 230 [M+ - 15; 22], 215 [M+ - 30; 8],
202 [M+ - 45; 15]. HRMS (EI): 245, exact mass found 245.1205,
calcd for C18H15N 245.120 450. Mp [4a]‚HCl: 178-181 °C.
(Z)-1-Cyano-2,3-bis(4-tert-butylphenyl)pent-1-en-3-yne. Com-
pound 3b was prepared by heating 17.1 mg of 1b in 3 mL of heptane
at 98 °C for 7 h. The solvent was removed, and the product was
purified by preparative TLC (hexane/ethyl acetate, 25/1). Yield: 13.3
mg, 78%. IR (CCl4): 2965 s, 2869 m, 2226 m, 2208 m, 1734 m, 1605
m, 1462 m, 1364 m, 1267 m, 1109 m, 1018 m, 838 m. 1H NMR (200
MHz, CDCl3): 7.26-7.12 (m, 8H), 2.18 (s, 3H), 1.29 (s, 9H), 1.28 (s,
9H). 13C NMR (100 MHz, CDCl3): 152.6, 152.1, 137.3, 133.0, 129.1,
129.06, 125.5, 125.2, 119.8, 117.0, 98.0, 80.5, 79.6, 34.75, 34.7, 31.2,
21.6, 20.7. MS (EI): 369 [M+; 6], 354 [M+ - 15; 11], 313 [M+
-
56; 7], 199 [M+ - 170; 25], 184 [M+ - 185; 100]. HRMS (EI): 369,
exact mass found 369.1389, calcd for C21H23NO3S 369.139 864.
Copper(I) (4-tert-Butylphenyl)acetylide.15 Nitrogen was bubbled
through 250 mL of a concentrated aqueous ammonia solution for 30
min, after which 10.7 g of CuI was added. An oxygen-free ethanolic
solution of 9.2 g of 4-tert-butylphenylacetylene was cannulated into
the well-stirred copper salt solution. A yellow precipitate formed. The
resulting suspension was filtered, and the yellow solid was washed
consecutively with 200 mL of water, 200 mL of ethanol, 200 mL of
hexane, 200 mL of ethanol, and 100 mL of water. After the mixture
was dried (0.01 Torr) at room temperature, 9.20 g of copper acetylide
was obtained. Yield: 9.20 g, 43.4 mmol, 80%.
1,6-Bis(4-tert-butylphenyl)-3-aza-4-methylhex-3-ene-1,5-
diyne.1,13 Compound 1b was prepared, starting with a suspension of
0.55 g of copper acetylide (2.6 mmol) in THF, to which was added a
solution of 0.91 g of 4-tert-butylphenylacetylene (5.8 mmol) and 3.2
mL of butyllithium solution (1.6 M in hexane) in THF at 0 °C. When
the mixture became clear, it was added dropwise at -28 °C to a solution
of 1.91 g of methyl (4-tert-butylphenyl)ethynyl ketoxime tosylate in
150 mL of THF. The mixture was allowed to warm to room
temperature over 12 h, and the reaction was quenched with saturated
aqueous NH4Cl. The mixture was diluted with ethyl acetate, the layers
were separated, and the organic solvents were removed. The residue
was suspended in a mixture of hexane/ethyl acetate, 4/1, and filtered
through silica. The resulting raw material was purified twice by column
chromatography: (1) 400 g of silica, 5.8 cm, hexane/ethyl acetate, 15/
1; (2) 600 g of silica, 5.8 cm, hexane/dichloromethane, 4/1 gradually
to 2/1, yielding a yellow solid consisting of the cis and trans isomers,
which were not separable by column chromatography or HPLC.
Yield: 0.250 g, 0.63 mmol, 12.2%. Mp: 126-128 °C. IR (CCl4):
2966 s, 2905 w, 2858 w, 2213 m, 2174 m, 1500 w, 1552 m. 1H NMR
(200 MHz, CDCl3): 7.53-7.35 (m, 8H), 2.50, 2.42 (2 s, 3H), 1.332,
1.327 (2 s, 18H). 13C NMR (125 MHz, CDCl3): 164.3, 160.9, 153.9,
153.6, 151.5, 151.3, 132.5, 132.4, 131.3, 131.1, 125.7, 125.6, 125.4,
125.35, 121.5, 121.3, 118.2, 118.1, 100.2, 100.15, 95.7, 92.6, 91.9,
89.9, 89.7, 86.0, 35.05, 35.0, 34.8, 31.21, 31.19, 31.1, 26.7, 25.5. MS
(EI): 355 [M+; 90], 340 [M+ - 15; 100], 284 [M+ - 71; 33], 242
[M+ - 113; 80]. Anal. Calcd for C26H29N (355.52): C, 87.84; H,
8.22; N, 3.94. Found: C, 88.11; H, 8.22; N, 4.00. UV (heptane):
195.5 (4.8), 251.5 (4.5), 335.5 (4.4).
31.1, 5.1. MS (EI): 355 [M+; 90], 340 [M+ - 15; 100], 284 [M+
-
71; 33], 242 [M+ - 113; 80]. HRMS (EI): 355, exact mass found
355.2299, calcd for C26H29N 355.230 000. UV (pentane): 202 (4.7),
239 (4.5), 324 (4.4).
(Z)-1-Cyano-2,3-diphenylpent-1-en-3-yne.1 Compound 3a was
prepared analogously to 3b by heating a sample of 1a in heptane at
120 °C for 6 h. After cooling, the solvent was removed. The residue
was purified by preparative TLC (hexane/ethyl acetate, 25/1). To
remove traces of 4a, the crude product was extracted five times with
10 mL of 0.1 N HCl and ethyl acetate. The organic layer was dried
and evaporated. The purity of 3a was checked by GC/MS. Analytical
data matched those in ref 1.
2-Iodopropane-d7. A 22.5 g portion of 2-propanol-d8 (d > 99%)
was added to 3.5 g of red phosphorus. The flask was fitted with a 25
mL dropping funnel filled with 42 g of iodine on top of which there
was a condenser. The alcohol was heated to reflux, and small portions
of the collected alcoholic iodine solution were dropped into the flask.
After 4 h, the addition was finished, and the mixture was heated to
reflux for 12 h. The product was purified by distillation through a 20
cm Vigreux column. Yield: 40.40 g, 228.2 mmol, 70%. Bp: 87.5
°C. IR (CCl4): 2236 m, 2212 m, 2122 m, 2059 m, 1148 m, 987 s,
977 s. 1H NMR (300 MHz, CDCl3): 4.3 (C2-H, I ) 14.7), 1.8 (C1-
H, I ) 28.9). MS: 177 [M+; 32.2, d7], 176 [M+; 0.75, d6], 159 [M+
1,6-Diphenyl-3-aza-4-methylhex-3-ene-1,5-diyne. Compound 1a
was prepared analogously to 1b and gave analytical data identical with
those reported in ref 1.
2-Methyl-3,4-diphenylpiperidine.16 4-Oxo-1,2-diphenylpentanecar-
bonitrile was prepared, following procedure of El Bouz et al.17 Then
5.0 g of the cyano ketone was dissolved in ethanol/THF and degassed
by sparging with argon. PtO2‚H2O (100 mg) was added to the solution,
which was kept under 8 atm of hydrogen for 17 h at 80 °C. The catalyst
was destroyed by adding dichloromethane, the mixture was filtered,
and the basic compounds were transferred to the water layer by adding
250 mL of 1 N HCl and ether. The water layer was reextracted with
500 mL of 1 N NaOH and ether. The organic layer was dried with
Na2SO4 and evaporated. An analytical sample was distilled (0.5 Torr,
180 °C). The sample contained all eight possible stereoisomers. 1H
NMR (200 MHz, CDCl3): 7.22-6.80 (m, 10H), 3.58-2.80 (m, 4H),
2.20-1.40 (m, 4H), 1.40-1.10 (4d, 3H).
- CD3; 0.77], 127 [I+; 16.6], 50 [M+ - I; 32.2, 100, d7], 49 [M+
-
I; 2.4, d6]. An isotopic purity of 99.7% was determined by GC/MS.
The deuteration grades at C1 (99.7%) and at C2 (99.5%) were
1
determined by H NMR.
Diisopropyl Ether-d14.19 Silver nitrate (23.5 g) was dissolved in
water from which silver hydroxide was precipitated with 200 mL of 5
N NaOH solution. The precipitate was washed with water and mixed
with 20 g of 2-iodopropane-d7. The mixture was refluxed for 5 h and
then distilled. Careful distillation over metallic sodium produced the
pure ether. Yield: 0.76 g, 13.1 mmol, 12%. Bp: 61 °C. IR (CCl4):
2-Methyl-4,5-diphenylpyridine.18 Compound 4a was prepared
from 2-methyl-3,4-diphenylpiperidine. Then 500 mg of the mixture
of the isomeric piperidines was mixed with 100 mg of Pd/C (10% Pd)
1
2228 s, 2147 m, 2094 m, 1214 m, 1159 s, 1089 s, 1048 s, 964 m. H
NMR (200 MHz, CDCl3): 3.55 (C2-H, I ) 24.5), 1.08 (C1-H, I )
44.3). MS: 116 [M+; 0.4, d14], 98 [M+ - CD3; 19.8, d11], 97 [M+
- CD3; 0.77, d10], 66 [M+ - C3D7; 7, d11], 58 [M2+], 50 [C3D7; 100],
49 [C3D6H; 4.78]. An overall isotopic purity of 99.7% was determined
(15) Hein, F.; Meyer, A. Z. Anal. Chem. 1927, 72, 30.
(16) Henecka, H. Chem. Ber. 1949, 82, 105.
(17) El Bouz, M.; Roux-Schmitt, M.-C.; Wartski, L. J. Chem. Soc. 1979,
779.
(18) Eliel, E. L.; McBride, R. T.; Kaufmann, S. J. Am. Chem. Soc. 1953,
75, 4291.
(19) Erlenmeyer, E. Ann. Chem. 1863, 126, 305. Glass, J. V. S.;
Hinshelwood, C. N. J. Chem. Soc. 1929, 1804.

Chemistry of the 2,5-Didehydropyridine Biradical
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 379
by GC/MS. The levels of deuteration at C1 (99.7%) and C2 (99.5%)
1
were determined by H NMR.
Bergman Cyclization Kinetics. A 100 µL portion of 1b in heptane
(0.14 mM) was pipetted into a 3 mL screw cap vial. The vial was
thermolyzed in a constant-temperature oil bath (Haake F3, (0.2 °C).
After thermolysis, the vial was cooled in water, and 20 µL was injected
into the HPLC. The substances 1b and 3b were monitored at a fixed
wavelength of 337 nm. The cyclization was investigated at five
different temperatures: 82.1, 92.0, 105.0, 114.5, and 124.4 °C. At each
temperature three to six samples were taken. The reaction rates were
determined by dividing the initial integrals of 1b by the obtained
integrals after a certain time. The 24 resulting points were fitted to
the Arrhenius equation, yielding a linear fit with a correlation of 0.996.
An activation energy of 23.1 ( 1.5 kcal/mol and an A-factor log10
A
of 9.7 ( 0.9 were obtained. This sample size requires a t-value of 2,
leading to the reported error bond of 2σ.
Trapping Procedure. A typical experiment was performed as
follows: Due to the hydrolysis of the azaenediyne at higher temper-
atures and low pH values, the reaction solutions were prepared in a
glovebox. A sample of 1a was freshly purified by preparative TLC
and checked by GC/MS prior to use. 1a (1.3 mg) was dissolved in
1.3 mL of distilled (Na, benzophenone) diisopropyl ether. Then 100
µL of this solution was pipetted into a 3 mL screw cap vial, to which
were added 100 µL of distilled (from CaH2) 2-fluoropyridine and 10
µL of a 0.1 M HBF4‚Et2O solution in diisopropyl ether with a few
drops of THF. The solution was thermolyzed at 97 °C for 31 min.
The solution was allowed to cool, and K2CO3 was added. The mixture
was filtered through a silica plug with ethyl acetate and concentrated
to about 100 µL prior to injection into the GC/MS for analysis.
Ab initio calculations were performed with Gaussian 9420 on either
IBM RS/6000-590 or DEC Alpha AXP 8400 5/300 workstations.
Energies along the ring-closure/ring-opening reaction coordinate for
the parent azaenediyne were computed at the CASSCF(6×6)/6-31G*
level for CASSCF(6×6)/3-21G-optimized geometries obtained in the
following fashion. The geometry specification was made such that
the CC and CN bond lengths for the two bonds being made or broken
could be fixed. Fixing both bond lengths to a large enough value led
to the localization of the six orbitals needed for the CASSCF (two
NBMO’s and two pairs of σ and σ* orbitals along the bonds being
made or broken). A localized active space along the reaction coordinate
was maintained by stepping the bond lengths back in small increments
while allowing the remaining degrees of freedom to relax. Once the
minima and transition states were located by this procedure, uncon-
strained optimizations to refine the geometries were performed. All
stationary points were then checked with frequency calculations to
confirm that they had either zero or one negative force constants.
Dynamic correlation effects were then included with the CASMP2
method. Comparable calculations21 for p-benzyne at a variety of levels
have appeared in the last several years.
Figure 1. Arrhenius plot of the logarithmic rate of disappearance of
1b versus inverse temperature. Twenty-four points at five temperatures
(82.0, 92.0, 105.0, 115.5, and 124.4 ( 0.2 °C) yield the activation
parameters: log10 A ) 9.7 ( 0.9 and Ea ) 23.1 ( 1.5 kcal/mol. Error
bounds are quoted as 2σ.
surface extends that comparison to the 2,5-didehydropyridine biradical.
In this case, abstraction by both “ends” of the biradical was considered,
the two “ends” no longer being degenerate. A 4 × 4 active space for
the CASSCF and CASMP2 calculations was found by setting the
distances from the hydrogen being abstracted to both the biradical and
the incipient hydroxylmethyl radical large. This led again to localization
of the four orbitals in the active space (the two biradical NBMO’s and
the methanol σCH and σ*CH). The two-dimensional potential surface
was then generated by stepping the two bond lengths over a fine grid
while simultaneously allowing all other degrees of freedom to relax
(no symmetry constraints). The points used to generate the final
contours were CASMP2/6-31G** single-point energies computed at
CASSCF/3-21G geometries with the 4 × 4 active space.
Results
The overall yield for the synthesis is low, but sufficient
quantities for mechanistic studies are nevertheless readily
obtained from inexpensive starting materials. Arrhenius activa-
tion parameters for the Bergman cyclization of 1b to the
6-methyl-3,4-bis(4-tert-butylphenyl)-2,5-didehydropyridine bi-
radical 2b were extracted from the linear fit shown in Figure 1,
giving log A ) 9.7 ( 0.9 and Ea ) 23.1 ( 1.5 kcal/mol with
the error bounds reported as 2σ. These figures are entirely
consistent with the earlier qualitative report for the cyclization
by David and Kerwin.1
A two-dimensional potential surface for the hydrogen abstraction
reaction between the biradical and methanol was obtained similarly.
The procedure is essentially identical with that used previously by
Logan and Chen22 to compare the abstraction of hydrogen from
methanol by p-benzyne to that by the phenyl radical. The present
For the right combination of conditions, listed in Table 1,
we observe the product of hydrogen abstraction reactions by
the didehydropyridine biradical 2a in trapping experiments.
Under each set of conditions, the GC/MS trace was searched
for the characteristic mass spectra of 2-methyl-4,5-diphenylpy-
ridine (4a), 1-cyano-1,2-diphenylpent-1-en-3-yne (3a), pheny-
lacetonitrile (5), and 1-phenylbut-1-yn-3-one (6), which are, as
noted previously, the products of hydrogen abstraction by an
intermediate biradical, the cyclization/ring-opening product, and
the two hydrolysis products of 1a.
In the absence of added acid, no hydrogen abstraction
products could be found in the 100 °C thermolysis of 1a. As
expected, 3a was formed in high yield, again in agreement with
the observations of David and Kerwin.1 With a small amount
of acid (<1 molar equiv relative to azaenediyne), there was
(20) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.;
Johnson, B. G.; Robb, M. A.; Cheeseman, J. R.; Keith, T.; Petersson, G.
A.; Montgomery, J. A.; Raghavachari, K.; Al-Laham, M. A.; Zakrzewski,
V. G.; Ortiz, J. V.; Foresman, J. B.; Cioslowski, J.; Stefanov, B. B.;
Nanayakkara, A.; Challacombe, M.; Peng, C. Y.; Ayala, P. Y.; Chen, W.;
Wong, M. W.; Andres, J. L.; Replogle, E. S.; Gomperts, R.; Martin, R. L.;
Fox, D. J.; Binkley, J. S.; Defrees, D. J.; Baker, J.; Stewart, J. P.; Head-
Gordon, M.; Gonzalez, C.; Pople, J. A. Gaussian 94, ReVision C.3; Gaussian,
Inc.: Pittsburgh, PA, 1995.
(21) Noell, J. O.; Newton, M. D. J. Am. Chem. Soc. 1979, 101, 51.
Wierschke, S. G.; Nash, J. J.; Squires, R. R. J. Am. Chem. Soc. 1993, 115,
11958. Nicolaides, A.; Borden, W. T. J. Am. Chem. Soc. 1993, 115, 11951.
Kraka, E.; Cremer, D. J. Am. Chem. Soc. 1994, 116, 4929. Lindh, R.;
Persson, B. J. J. Am. Chem. Soc. 1994, 116, 4963. Lindh, R.; Lee, T. J.;
Bernhardsson, A.; Persson, B. J.; Kalstro¨m, G. J. Am. Chem. Soc. 1995,
117, 7186. Nash, J. J.; Squires, R. R. J. Am. Chem. Soc. 1996, 118, 11872.
Lindh, R.; Schu¨tz, M. Chem. Phys. Lett. 1996, 258, 409. Cramer, C. J.;
Nash, J. J.; Squires, R. R. Chem. Phys. Lett., submitted.
(22) Logan, C. F.; Chen, P. J. Am. Chem. Soc. 1996, 118, 2113.

380 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Hoffner et al.
Table 1. Representative Trapping Results in the Thermolysis of 1a under a Variety of Conditionsa
HBF4‚Et2O
identifiable
products by
GC/MS
N
F
(100 mM in
(buffer)
substrate
diisopropyl ether)
conditions
1a, 1.0 mg/mL
1a, 1.0 mg/mL
100 µL
100 µL
100 µL (25 equiv)
10 µL (2.5 equiv)
100 °C/1 h
100 °C/1 h
-
3a, 4a, 5a, 6a +
other products
1a, 1.0 mg/mL
1a, 1.0 mg/mL
1a, 1.0 mg/mL
1a, 1.0 mg/mL
100 µL
100 µL
100 µL
-
1 µL (0.25 equiv)
-
10 µL (2.5 equiv)
10 µL (2.5 equiv)
100 °C/1 h
100 °C/1 h
25 °C/1 h
100 °C/1 h
3a
3a
3a
traces of several
products
3a, 1.0 mg/mL
4a, 1.0 mg/mL
4a, 1.0 mg/mL
100 µL
100 µL
100 µL
10 µL (2.5 equiv)
10 µL (2.5 equiv)
100 µL (25 equiv)
100 °C/1 h
100 °C/1 h
100 °C/1 h
3a
4a
-
a Because of the presumed isomerization of 1a to 3a in the GC/MS injector port, any residual 1a appears as 3a on the column. The donor solvent,
DIPE, is diisopropyl ether (100 µL).
Table 2. Absolute Energies and Singlet-Triplet Splittings from ab Initio Calculations at the CASSCF and CASMP2 Levels for the
p-Benzyne, 2,5-Didehydropyridine, and 2,5-Didehydropyridinium Biradicals
•
•
•
+
N
H
N
•
•
•
CASSCF(6×6)6-31G* singlet (hartrees)
CASSCF(6×6)6-31G* triplet, vertical
CASSCF(6×6)6-31G* triplet, adiabatic
CASSCF(6×6)6-31G* ∆EST (kcal/mol)
-229.4235532
-229.4209942
-229.4217876
1.6 (vertical)
1.1 (adiabatic)
-230.0963386
-230.0929887
-230.0952040
2.1 (vertical)
-245.4234846
-245.4134308
-245.4170952
6.3 (vertical)
4.0 (adiabatic)
-246.1206479
-246.1075771
-246.1135411
8.2 (vertical)
-245.7745936
-245.7710849
-245.7720467
2.2 (vertical)
1.6 (adiabatic)
-246.4582237
-246.4540073
-246.4568636
2.7 (vertical)
CASMP2(6×6)6-31G* singlet (hartrees)
CASMP2(6×6)6-31G* triplet, vertical
CASMP2(6×6)6-31G* triplet, adiabatic
CASMP2(6×6)6-31G* ∆EST (kcal/mol)
0.7 (adiabatic)
4.5 (adiabatic)
0.9 (adiabatic)
little change. Although very large excesses of acid (>10 equiv)
destroyed 1a as well as 4a, the thermolysis of 1a with
intermediate amounts of acid (e.g. 2.5 equiv) resulted, aside
from the production of 3a, in the reproducible production of
the trapping product, 2-methyl-4,5-diphenylpyridine (4a) in
small, but clearly detectable, quantities. A rough estimate of
the combined yield for 3a and 4a is the order of a percent, with
a 4a/3a ratio of ∼1:20. A large amount of hydrolysis products,
plus polymeric material, accounted for most of the material. If
the 2-fluoropyridine buffer was omitted, the overall yield of
identifiable products dropped to the limits of detection by GC/
MS, which presumably happens because the strong acid
protonates the azaenediyne itself under those conditions, leading
to irreversible loss of material prior to cyclization. A search
for 4a-d2 in the same reaction conducted with perdeuterated
diisopropyl ether was inconclusive. The results reported here
comprise the first observation of a hydrogen abstraction reaction
by a 2,5-didehydropyridine-type biradical. The conditions under
which hydrogen abstraction occurs had been predicted, as
discussed below, on the basis of ab initio calculations and a
simple model for biradical reactivity.
solutions in hydrogen-containing solvents, for periods of 6
months showed traces of 4a, according to GC/MS analysis.
Ab initio calculations were performed to probe three proper-
ties of the title biradical. The vertical and adiabatic singlet-
triplet gaps of the 2,5-didehydropyridine biradical, the 2,5-
didehydropyridinium biradical, and p-benzyne (as a reference)
were found at the CASSCF(6×6)/6-31G* and CASMP2(6×6)/
6-31G* levels and are listed in Table 2.
An experimentally measured value for p-benzyne23 of 3.8 (
0.5 kcal shows that the present calculations slightly overestimate
the stability of the metastable triplet relative to the singlet ground
state. Nevertheless, the trend is clear that incorporation of a
nitrogen into the ring increases the singlet-triplet splitting of
the biradical by several kcal/mol relative to p-benzyne at the
same level of theory. Moreover, protonation of that nitrogen
reduces the singlet-triplet splitting back to a value comparable
to that in the hydrocarbon system.10
One-dimensional relaxed potential surfaces for the tandem
Bergman cyclization/ring-opening reaction of the all-hydrocar-
bon system (for reference), the parent azaenediyne, and its
N-protonated version are shown in Figures 2-4.
Control experiments, in which either 3a or 4a were prepared
in the same solution for which hydrogen abstraction products
were observed, and then thermolyzed at 100 °C for 1 h prior to
GC/MS analysis, found both products to be stable to the reaction
conditions. Importantly, the ring-opening product, 1-cyano-1,2-
diphenylpent-1-en-3-yne (3a), under the reaction conditions,
gave rise to neither 4a nor the range of other products from 1a,
establishing that the key pyridine product could not have arisen
from an acid-induced rearrangement of the ring-opening product.
Interestingly, samples of 1a, stored at 4 or 25 °C as dilute
The reference surface for hex-3-ene-1,5-diyne in Figure 2
agrees reasonably well with the original Bergman estimate24
and recent experimental redeterminations by Roth25 and Squires.26
The CASMP2 calculation clearly places the biradical too low
in energy, relative to its closed-shell precursor; the CASSCF
(23) Wenthold, P. G.; Hu, J.; Squires, R. R.; Lineberger, W. C. J. Am.
Chem. Soc., submitted.
(24) Bergman, R. G. Acc. Chem. Res 1973, 6, 25.
(25) Roth, W. R.; Hopf, H.; Horn, C. Chem. Ber. 1994, 127, 1765.
(26) Wenthold, P. G.; Squires, R. R. J. Am. Chem. Soc. 1994, 116, 6961.

Chemistry of the 2,5-Didehydropyridine Biradical
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 381
Figure 2. Computed one-dimensional reaction coordinate for the parent
Bergman cyclization. Energy differences in kcal/mol at the CASSCF-
(6×6)/6-31G* and CASMP2(6×6)/6-31G* (numbers in brackets) levels
of theory.
Figure 5. Computed contour plot of the two-dimensional surface for
the hydrogen abstraction reaction of 2,5-didehydropyridine and metha-
nol in the region around the transition state. The two coordinates are
the respective C-H bond lengths from the abstracted hydrogen to the
biradical and the donor. The energies along the contours are given in
kcal/mol relative to the asymptotic limit of large separation (10 Å) in
the reactant channel (not shown). The transition vector is indicated,
with the transition state at the crossbar.
protonation hardly affects the activation barrier for the initial
Bergman cyclization, but markedly affects the barrier for ring
opening of the intermediate biradical. While the precise
numbers will probably change somewhat at a higher level of
theory, the trend, as discussed below, is unlikely to be altered
in any significant way.
A two-dimensional potential surface for hydrogen abstraction
from methanol by an unsubstituted 2,5-didehydropyridine is
shown in Figure 5.
Figure 3. Computed one-dimensional reaction coordinate for the
Bergman cyclization of an unsubstituted azaenediyne. Energy differ-
ences in kcal/mol at the CASSCF(6×6)/6-31G* and CASMP2(6×6)/
6-31G* (numbers in brackets) levels of theory.
The two coordinates are the distance from the hydrogen atom
to carbons on either the biradical or the incipient hydroxymethyl
radical. The transition state is slightly earlier than for the
comparable reaction by p-benzyne, occurring when the bond
lengths from the transferring hydrogen to the biradical, and the
hydroxymethyl moiety, are 1.45 and 1.28 Å, as compared to
1.375 and 1.275 Å for p-benzyne.22 The computed activation
energy is 13.8 kcal/mol, with the approach of the methanol to
the end of the didehydropyridine away from the nitrogen. The
4 × 4 active space at the transition-state geometry is shown in
Figure 6.
Approach to the other end gave a slightly higher activation
energy and was therefore not considered further. The activation
energy, in this case, is 4.3 kcal/mol higher than that computed
for the same reaction of p-benzyne at the same level of theory.22
Computation of potential surfaces for the 2,5-didehydropyridium
biradical was less successful. We found the ion-dipole
interaction27 to be so large that it completely dominated the
approach of the protonated biradical to the methanol moiety.
In solution, the purely electrostatic interaction would be largely
screened out by the solvent dielectric, which unfortunately is
Figure 4. Computed one-dimensional reaction coordinate for the
Bergman cyclization of an unsubstituted protonated azaenediyne. Energy
differences in kcal/mol at the CASSCF(6×6)/6-31G* and CASMP2-
(6×6)/6-31G* (numbers in brackets) levels of theory.
calculation does very well with the biradical, but overestimates
the activation energy by about 10 kcal/mol. This systematic
errorsreasonable ∆E but Ea too highspresumably carries over
into the didehydropyridine surfaces as well. The surfaces for
the unprotonated and protonated forms of the didehydropyridine
biradical are strongly asymmetric, with the product nitrile lying
substantially below the initial azaenediyne. Significantly,
(27) Farneth, W. E.; Brauman, J. I. J. Am. Chem. Soc. 1976, 98, 7891.
Olmstead, W. N.; Brauman, J. I. J. Am. Chem. Soc. 1977, 99, 4219.
Asubiojo, O. I.; Brauman, J. I. J. Am. Chem. Soc. 1979, 101, 3715. Pellerite,
M. J.; Brauman, J. I. J. Am. Chem. Soc. 1980, 102, 5993. Sharma, S.;
Kebarle, P. J. Am. Chem. Soc. 1982, 104, 19. Caldwell, G.; Magnera, T.
F.; Kebarle, P. J. Am. Chem. Soc. 1984, 106, 959. McMahon, T. B.; Heinis,
T.; Nicol, G.; Hovey, J. K.; Kebarle, P. J. Am. Chem. Soc. 1988, 110, 7951.

382 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Hoffner et al.
Figure 6. Orbitals in the 4 × 4 active space for the CASSCF and CASMP2 calculations of the hydrogen abstraction reaction by the
2,5-didehydropyridine biradical and methanol.
not included in the present computational model. Without
explicit inclusion of solvent, presumably through a SCRF
procedure,28 the surfaces for the ionic reaction are skewed
beyond usefulness and are therefore not further considered in
this study.
Geometries for all stationary points along the one-dimensional
potential surfaces, as well as absolute energies for the single-
point ab initio calculations that go into the two-dimensional
surfaces, are included in the Supporting Information.
biradical has a singlet ground state. For a singlet ground-state
1,4-arene biradical to engage in radical-like chemistry, e.g.
hydrogen abstraction, it must go back through a curve-crossing
on the potential surface which adds an extra increment in the
activation energy that scales with the singlet-triplet gap. To
state the same result in another way, the singlet lies below the
triplet because it is stabilized, not because the triplet is
destabilized. Accordingly, one must pay back that stabilization
energy to reach a transition state where the two electrons in the
NBMO’s have been effectively uncoupled. The particular
mechanism for stabilization (relative to additivity or homodes-
motic reaction estimates) for the present 1,4-arene biradicals is
through-bond coupling between the two nominal NBMO’s, as
has been discussed extensively by the groups of Hoffmann30
and Paddon-Row.31 The p-benzyne biradical was, in fact, one
of the original examples presented by Hoffmann to demonstrate
the effect. In a further step, one can surmise that an increase
in the electron density in the intervening σ-bonds could increase
the through-bond coupling and hence the singlet-triplet split-
ting. Conversely, decreased electron density could decrease the
coupling and reduce the singlet-triplet splitting.10 The in-plane
lone pair of the 2,5-didehydropyridine biradical lies antiperipla-
nar to the σ-bonds coupling the NBMO’s and therefore could
donate electron density. However, when the nitrogen is
protonated, the effect is reversed. This qualitative reasoning is
confirmed by the ab initio computed singlet-triplet gaps, where
the didehydropyridine, with its lone pair, shows a much larger
singlet-triplet gap than p-benzyne. One infers, then, that the
unprotonated 2,5-didehydropyridine biradical should abstract
hydrogens much more slowly than p-benzyne and more slowly
than the protonated species. The ab initio calculations support
this contention as well, with the increase in the computed
activation energy for the didehydropyridine biradical com-
mensurate with the increase in singlet-triplet splitting. Pro-
tonation, on the other hand, should both reduce the singlet-
triplet splitting and restore the rate of hydrogen abstraction to
almost the original p-benzyne value. The absolute values for
the computed Ea for hydrogen abstraction appear to be slightly
Discussion
The ready availability of the C,N-dialkynylimines, albeit in
low overall yield, makes mechanistic studies feasible. As had
been reported by David and Kerwin,1 the compounds show a
half-life of ∼40-50 min at 110 °C, presumably undergoing ring
closure/ring opening in rapid succession to generate the observed
nitrile product. Attempts reported in that work to trap an
intermediate biradical, by either hydrogen transfer from a
suitable donor, or by intramolecular addition to an unsaturated
moiety, were unsuccessful, leading to the suggestion that perhaps
the biradical was not on the reaction coordinate at all.
The present report constitutes the continuation of the ongoing
project10-12,22 whereby we have explored structural effects on
the hydrogen abstraction rates by singlet biradicals. Extending
a qualitative thermochemical model for predicting the heats of
formation for singlet carbenes and biradicals,11 we had predicted
that singlet biradicals would abstract hydrogens more slowly
than comparable monoradicals. The prediction was experimen-
tally confirmed for the 9,10-didehydroanthracene biradical,12 for
which the hydrogen abstraction rate from 2-propanol was found
to be reduced by 2-3 orders of magnitude relative to the phenyl
or 9-anthryl radicals. A comparable rate reduction was subse-
quently found for the parent p-benzyne biradical,29 as well, in
a trapping study of its hydrogen abstraction reactions with
methanol. Briefly stated, the model predicts that a “noninter-
acting” biradical is best approximated by the triplet, even if the
(28) Wong, M. W.; Frisch, M. J.; Wiberg, K. B. J. Am. Chem. Soc. 1991,
113, 4776. Wong, M. W.; Wiberg, K. B.; Frisch, M. J. J. Am. Chem. Soc.
1992, 114, 523. Wong, M. W.; Wiberg, K. B.; Frisch, M. J. J. Chem. Phys.
1991, 95, 8991. Wong, M. W.; Wiberg, K. B.; Frisch, M. J. J. Am. Chem.
Soc. 1992, 114, 1645. Miertus, S.; Tomasi, J. Chem. Phys. 1982, 65, 239.
Miertus, S.; Scrocco, E.; Tomasi, J. Chem. Phys. 1981, 55, 117.
(29) Roth, W. R.; Hopf, H.; Wasser, T.; Zimmermann, H.; Werner, C.
Liebigs Ann. 1996, 1691.
(30) Hoffmann, R.; Iwamura, A.; Hehre, W. J. J. Am. Chem. Soc. 1968,
90, 1499. Hoffmann, R. Acc. Chem. Res 1971, 4, 1.
(31) A recent review of experimental and theoretical work on through-
bond coupling may be found in the following: Paddon-Row, M. N.; Jordan,
K. D. In Modern Models of Bonding and Delocalization; Liebman, J. F.,
Greenberg, A., Eds.; VCH Publishers: New York, 1988; Chapter 3.

Chemistry of the 2,5-Didehydropyridine Biradical
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 383
Scheme 3
high, but the difference in activation energies implied by the
computed singlet-triplet gaps suggests, nevertheless, a rate
difference of perhaps 2 orders of magnitude, triggered by
protonation. This should appear as a pH-dependence for the
hydrogen abstraction products obtained by thermolysis of the
azaenediyne precursors to the biradicals.
pKa for the conjugate acid would be ∼5.5-6.5. On the basis
of the known basicity of N-alkylimines,32 and the known base-
weakening effects of electronegative N-substituents (heteroat-
oms, sp2- or sp-hybridized carbon) in the analogous amidines,
the azaenediyne should be a much weaker base than the biradical
by an estimated 4-5 pKa units. The nitrile should be the
weakest base of all the structures. Upon protonation, the less
basic compounds will be raised in energy relative to the more
basic ones, which is exactly what is depicted in Figure 4. The
transition states should also move in the direction predicted by
the Hammond postulate. The net result is that the barrier for
the irreversible deactivation of the biradical by a retro-Bergman
reaction is substantially raised by protonation.
A quick examination of the trapping results confirms the
prediction that the hydrogen abstraction reaction product, i.e.,
2-methyl-4,5-diphenylpyridine (4a), is formed when the aza-
enediyne 1a is thermolyzed in the presence of a moderate excess
of acid. The low combined yield of 3a and 4a in the acid-
containing runs can be assigned, based on the control experi-
ments, to the irreversible loss of 1a to hydrolysis or polymer-
ization in competition with Bergman cyclization. The important
experimental observable here, however, is not the absolute yield
of products but rather the ratio of 4a to 3a, which is zero (no
4a) in the absence of added acid and approximately 1:20 for
the runs with 2.5 equiv of HBF4. The absence of any trace of
hydrogen abstraction by the unprotonated azaenediyne in the
100 °C thermolyses is conveniently explained by two separate
effects, the relative magnitudes of which remain yet to be
investigated. The first of the two is the effect of protonation
upon hydrogen abstraction rate by the biradical, as discussed
above. Working in the same direction is the second effect, a
lifetime effect, which can be understood by comparing Figures
3 and 4. Whereas the parent hydrocarbon undergoes a degener-
ate Bergman cyclization reversibly, the “tilted” potential surface
for the azaenediyne makes the retro-Bergman ring opening to
the nitrile irreversible with an activation barrier of only ∼7 kcal/
mol. The decreased macroscopic half-life for the azaenediyne
relative to the parent hydrocarbon, commented upon by David
and Kerwin,1 arises because the disappearance of the azaene-
diyne reflects the kinetics of rate-limiting cyclization, while that
of the original Bergman enediyne reflects the rate of bimolecular
reactions of the biradical. Examination of Figure 4 shows,
however, that protonation dramatically increases the ring-
opening barrier for the biradical. The altered energetic ordering
of the azaenediyne, the biradical, and the product nitrile upon
protonation is, in hindsight, obvious. One should note that the
basicity of the 2,5-didehydropyridine biradical should not be
too different from that of the comparable pyridine, meaning that
To manipulate the extent of hydrogen abstraction by modulat-
ing pH, one needs to explicitly consider the proton-transfer
equilibria. It would be a gross oversimplification to say that
one simply needs to add acid to protonate the biradical and thus
suppress ring opening. Proton-transfer reactions between elec-
tronegative heteroatoms in a hydrogen-bonding solvent are
exceedingly fast, with exothermic proton transfers occurring at
or near the diffusion limit.33 At any given pH, deprotonation
of the 2,5-didehydropyridinium biradical (itself slow to ring
open) in a fast preequilibrium would lead to a irreversible decay
via the fast ring opening of the unprotonated biradical. The
observed rate for ring opening for the protonated biradical would
be, therefore, that for the unprotonated species, weighted by
the equilibrium fraction of unprotonated biradical (Scheme 3).
This leads to an estimate for the conditions under which
hydrogen abstraction by 2,5-didehydropyridine can occur, in
competition with ring opening. For argument’s sake, take
Arrhenius parameters for ring-opening of the unprotonated
biradical to be A ) 1012 s-1 and Ea ) 7 kcal/mol. For hydrogen
abstraction by the unprotonated biradical, take A ) 108 M-1
s-1 and Ea ) 6 kcal/mol. For the protonated biradical, assume
the same preexponential, but a lower activation energy, Ea ) 4
(32) Cordes, E. H.; Jencks, W. P. J. Am. Chem. Soc. 1963, 85, 2843.
(33) Proton-transfer reactions in aqueous solution between nitrogen or
oxygen bases are very fast, even diffusion-controlled, when the pKa
difference is not too large. Protonation should therefore be faster than
hydrogen abstraction by the biradical. Maskill, H. The Physical Basis of
Organic Chemistry; Oxford University Press: Oxford, 1985.

384 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Hoffner et al.
kcal/mol,34 for hydrogen abstraction. One should note here that
that observation of traces of hydrogen abstraction product 4a
in long-term-stored samples of 1a kept at 4 °C (t1/2 ) 51 months
for Bergman cyclization) or 25 °C (t1/2 ) 3.6 months) suggests
that both Ea and A are lower for hydrogen abstraction by the
biradical than for ring opening. In other words, hydrogen
abstraction is favored enthalpically but disfavored entropically,
which is just what reasonable guesses based on literature
precedents suggest.34 Now, assume the ∆pKa between the
didehydropyridinium and a buffer (in this case 2-fluoropyri-
dine35) to be 6 units, with respective concentrations of biradical
(total protonated and unprotonated), buffer, and hydrogen donor
solvent at 10-5, 5, and 5 M, respectively, and a temperature of
100 °C. As noted above, we ignore the rate for direct ring
opening of the protonated biradical in this “straw-man” calcula-
tion. As one increases the amount of added (strong) acid from
zero to 0.25 equiv (relative to biradical) to 2.5 equiv, and on to
25 equiv, the ratio of protonated to unprotonated biradical goes
from 0 to 0.19 to 3.5 to 48. Accordingly, the ratio of ring-
opening product to hydrogen abstraction product falls from 523
to 141 to 10 to 0.75, with the qualitative effects of higher
abstraction rate by the protonated biradical, and its higher steady-
state concentration, both contributing comparably to the overall
effect. This rough estimate is completely consistent with the
experimental observations. One can rationalize the difficulty
in observing deuterium abstraction by considering that the
abstraction reaction, with its large primary kinetic isotope effect,
must compete against a ring opening that has no isotope effect.
Reduction of the ratio of abstraction product to ring-opening
product by perhaps 1 order of magnitude would make unam-
biguous detection problematic.
Scheme 4
types37-40 from 7.2 to between 6.2 and 6.6. More striking is
the effect of hyperglycemia, which has been shown to cause a
further acidification41 to a remarkably low intracellular pH of
5.5. Moderate hypoxia also contributes in the same direction.
In 1989, Tannock42 suggested selective acidification as a means
to target tumor cells, but the chemical arsenal lacked the
necessary tools needed for the construction of a pH-dependent
cytotoxin. Since then, there have been some attempts, notably
by Tietze43 and Shibuya44 and their co-workers, who synthesized
acid-labile prodrugs which hydrolyze rapidly at pH 6.2 to yield
toxic products. Even more promising is the very recent report,
by Vol’pin and co-workers,45 of a pH-triggered source of
reactive free radicals (under physiological conditions) modeled
on the alkylcobalamins (Scheme 4).
The rate of alkyl radical release from the alkylcobalt(III)
complex showed an sigmoidal dependence on pH between 4.0
and 8.0 with a (modest) 3-fold increase in the rate of radical
production as the pH dropped from 7.5 to 6.5 (aqueous
phosphate buffer, 20 °C). The chemical studies were im-
mediately followed by trials of the cobalt complex in rats with
Guerin carcinoma, mice with Ascit-L1210 leukemia, and rats
with sarcoma 45. While details differed in each case, the overall
trend was surprisingly good. Administration of between 20 and
100 mg/kg of the cobalt complex, accompanied by glucose
infusion to induce acidification (reported to be 0.90 pH units),
resulted in large tumor growth delays and mean increases in
life span from 23 to 61%. Complete regression was observed
in up to 40% of the test animals. The effect in the absence of
hyperglycemia was much less pronounced. The complex also
significantly enhanced the effectiveness of both radiotherapy
and chemotherapeutic agents such as cis-platin, as evidenced
by further increases in life expectancy.
How does this become a concept for rational drug design?
There have been several modified enediynes reported, where a
pendent group enhances molecular recognition or provides a
trigger for Bergman cyclization36 to improve selectivity for
tumor cells. However, none of the previous approaches sought
to tailor the actual hydrogen abstraction rates to improve
selectivity. Greater toxicity of a modified enediyne compound
in tumor cells versus normal cells might be achievable if
intracellular pH could be used as a trigger for increased reactivity
of a biradical. A microphysiological handle for the selection
of tumor over normal cells derives from the demonstrated ability
of hyperglycemia and/or certain existing drugs, e.g., amiloride,
nigericin, and hydralazine, to preferentially lower the intracel-
lular pH in tumor cells. Administration of amiloride or nigericin
at dosages where no discernible effect is found in normal cells
drops the intracellular pH in a wide range of tumor cell
The azaenediynes 1a and 1b, and the corresponding 2,5-
didehydropyridine biradicals 2a and 2b, prepared for this
(37) Adams, G. E.; Stratford, I. J. Int. J. Radiat. Oncol. Biol. Phys. 1994,
29, 231. Priyadarsini, K. I.; Dennis, M. F.; Naylor, M. A.; Stratford, M. R.
L.; Wardman, P. J. Am. Chem. Soc. 1996, 118, 5648 and references therein.
(38) Stubbs, M.; Rodrigues, L.; Howe, F. A.; Wang, J.; Jeong, K. S.;
Veech, R. L.; Griffiths, J. R. Cancer Res. 1994, 54, 4011.
(39) Lyons, J. C.; Ross, B. D.; Song, C. W. Int. J. Radiat. Oncol. Biol.
Phys. 1992, 25, 95. Song, C. W.; Lyons, J. C.; Griffin, R. J.; Makepeace,
C. M. Radiother. Oncol. 1993, 27, 252. Song, C. W.; Lyons, J. C.; Griffin,
R. J.; Makepeace, C. M.; Cragoe, E. J., Jr. Cancer Res. 1993, 53, 1599.
Song, C. W.; Lyons, J. C.; Makepeace, C. M.; Griffin, R. J.; Cragoe, E. J.,
Jr. Int. J. Radiat. Oncol. Biol. Phys. 1994, 30, 133. Song, C. W.; Kim, G.
E.; Lyons, J. C.; Makepeace, C. M.; Griffin, R. J.; Rao, G. H.; Cragoe, E.
J., Jr. Int. J. Radiat. Oncol. Biol. Phys. 1994, 30, 1161. Lyons, J. C.; Song,
C. W. Radiat. Res. 1995, 141, 216.
(40) Newell, K.; Wood, P.; Stratford, I.; Tannock, I. Br. J. Cancer 1992,
66, 311. Hasuda, K.; Lee, C.; Tannock, I. F. Oncol. Res. 1994, 6, 259.
(41) von Ardenne, M. AdV. Pharmacol. 1972, 10, 339. Osinsky, S.;
Bubnovskaya, L. Arch. Geschwulstforsch. 1984, 54, 463.
(42) Tannock, I. F.; Rotin, D. Cancer Res. 1989, 49, 4373.
(43) Tietze, L. F.; Fischer-Beller, A. Carbohydr. Res. 1994, 254, 169
and references therein.
(34) The assumed log A and Ea for both the Bergman cyclization and
the hydrogen abstractions are reasonable compared to the measured values
for the hydrocarbon cases in ref 29. These “straw-man” values do not
necessarily match the ab initio calculated ones, but are meant to represent
a reasonable guess at the real values given the systematic errors in the
computed values. In addition, the activation energy for hydrogen abstraction
from diisopropyl ether is undoubtedly lower than that from methanol by a
few kcal/mol.
(35) The pKa of 2-fluoropyridium has been variously reported to be
between -0.44 and 1.41, depending on the solvent. The more positive value
pertains to nonaqueous solution. Brown, H. C.; McDaniel, D. H. J. Am.
Chem. Soc. 1955, 77, 3752. Clark, H. R.; Beth, L. D.; Burton, R. M.; Garrett,
D. L.; Miller, A. L.; Muscio, O. J., Jr. J. Org. Chem. 1981, 46, 4363.
Abraham, M. H.; Duce, P. P.; Prior, D. V.; Barratt, D. G.; Morris, J. J.;
Taylor, P. J. J. Chem. Soc., Perkin Trans. 2 1989, 1355.
(44) Naoe, Y.; Kikuishi, J.; Ishigaki, K.; Iitsuka, H.; Nemoto, H.; Shibuya,
M. Tetrahedron Lett. 1995, 9165.
(45) Vol’pin, M.; Levitin, I.; Osinsky, S. Angew. Chem. 1996, 108, 2516.
Levitin, I. Ya.; Belkov, V. M.; Novodarova, G. N.; Shabarova, Z. A.;
Vol’pin, M. E. MendeleeV Commun. 1996, 153.
(36) Lu, Y.-F.; Harwig, C. W.; Fallis, A. G. J. Org. Chem. 1993, 58,
4202. Wang, J.; DeClercq, P. J. Angew. Chem. 1995, 107, 1898. Banfi, L.;
Guanti, G. Angew. Chem. 1995, 107, 2613. McPhee, M. M.; Kerwin, S.
M. J. Org. Chem. 1996, 61, 9385.

Chemistry of the 2,5-Didehydropyridine Biradical
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 385
Scheme 5
which can go into in vivo testing. If the high cytotoxicity of
the natural products could be preferentially displayed in tumor
cells, then one might achieve good therapeutic results with less
severe side effects than has been seen for natural enediyne
compounds. This work is underway.
Conclusion
We report the experimental observation of hydrogen abstrac-
tion by a 2,5-didehydropyridine biradical, formed by Bergman
cyclization of an azaenediyne. The qualitative reactivity of the
intermediate biradical agrees with theoretical predictions, and
is furthermore tunable by protonation. The enhanced yield of
hydrogen abstraction products at lower pH would be a potential
basis for the construction of an antitumor agent based on radical-
induced damage to cellular components by the title biradical.
mechanistic study, are not the ideal candidates for drug
development, although their chemistry does lay out the condi-
tions that a real candidate would have to fulfill. The azaene-
diynes 1a and 1b themselves are too labile with respect to
hydrolysis, and the pKa values for the didehydropyridinium
biradicals need to be adjusted by substitution. Moreover,
without inclusion of the azaenediyne moiety into a strained
macrocycle46 to lower the activation energy for Bergman
cyclization (without much affecting the retro-Bergman and
hydrogen abstraction), the temperatures needed for biradical
formation are 50-70 °C higher than physiological. Promising
structures under study include those shown in Scheme 5.
Nevertheless, with the conditions defined, the challenge now
becomes the selection and synthesis of a biradical precursor
Supporting Information Available: Geometries for all
stationary points along the one-dimensional potential surfaces
and absolute energies for the single-point ab initio calculations
that go into the two-dimensional surfaces (48 pages). See any
current masthead page for ordering and Internet access
instructions.
(46) Nicolaou, K. C.; Smith, A. L. Acc. Chem. Res. 1992, 25, 497.
JA9730223