Reactions of Charged Substrates. 5. The Solvolysis and Sodium Azide Substitution Reactions of Benzylpyridinium Ions in Deuterium Oxide

Article abstract of DOI:10.1021/jo960729j

Second-order rate constants and activation values were measured for the reactions with NaN3 of a series of 4-Y-substituted (Y = MeO, Me, H, Cl, and NO2) benzyl 3'-Z-substituted (Z = CN, CONH2, H, F, Ac) pyridinium chlorides in deuterium oxide. 3'-Cyanopyridine substrates reacted much faster than nicotinamide and pyridine substrates; in the pyridine series the 4-me, 4-H, and 4-Cl benzyl analogs did not react for up to 6 months at 96 deg C in 1.7 M NaN3.The 3'-cyanopyridine substrates do not exhibit borderline kinetic behavior, but the nicotinamide substrates do.The Hammett plot is flat for the NaN3 reaction of 3'-cyanopyridine substrates and increasingly V-shaped for the nicotinamide and pyridine substrates. the values of β_LG (four-point plot) for the NaN3 reaction of the 4-MeO benzyl substrates is -1.45, which is usually interpreted as being a very "late" activated complex.Two-pint Broensted "plots" for the other benzyl derivatives and for two N-methylpyridinium ions give values of β_LG in the same range.The second-order rate constant and activation values for N-methyl-3'-cyanopyridinium iodide are within the same range as those for the benzyl substrates.For the hydrolysis reaction, the Hammett plot is linear for 3'-cyanopyridine substrates (ρ' = -1.24) and flat for the nicotinamide substrates.The extent of hydrolysis of 0.005-0.05 M solutions of the 3'-cyanopyridine substrates depended on the initial concentration of substrate, and hydrolysis was slowed significantly or stopped completely in the presence of exogenous 3-cyanopyridine.These results show that an equilibrium is established among the products for the 4-MeO, 4-Me, 4-H, and 4-Cl substrates; the 4-NO2 substrate reacted too slowly to discern any difference.Data for the extent of hydrolysis were fitted by an equation derived assuming the equilibrium.Despite this limitation on a classic test of mechanism, the rates and ρ values are consistent with direct displacement by solvent and not with a unimolecular process.These results, which are rationalized in terms of the Pross-Snaik model, suggest that there are no ion-dipole complex intermediates in the benzyl series and show that borderline kinetic behavior is a function of leaving group ability and is not necessarily related to a change in mechanism.A computational approach was used to evaluate anomalous β_LG values for the hydrolysis and nucleophilic substitution reactions of the methylpyridinium ion substrates.It was found that neither the Nu-substrate bond lengths nor the difference in charge matched the β_LG values.The value if ΔΔS^activ. of -15 gibbs/mol between (4-methoxybenzyl)-3'-cyanopyridinium chloride and the corresponding dimethylsulfonium chloride in the NaN3 reaction, which is the result of the solvation of the pyridine at the transition state and the lack of solvation of SMe2, is used to argue that the source of NAD⁺ glycohydrolase "catalysis" of NAD⁺ bond cleavage is the result of desolvation of the leaving group upon binding.

Full text of DOI:10.1021/jo960729j

7360
J . Org. Chem. 1996, 61, 7360-7372
Rea ction s of Ch a r ged Su bstr a tes. 5. Th e Solvolysis a n d Sod iu m
Azid e Su bstitu tion Rea ction s of Ben zylp yr id in iu m Ion s in
Deu ter iu m Oxid e
Neil Buckley* and Norman J . Oppenheimer*
The Department of Pharmaceutical Chemistry, S-926, Box 0446, The University of California,
San Francisco, California 94143-0446
Received April 22, 1996X
Second-order rate constants and activation values were measured for the reactions with NaN3 of
a series of 4-Y-substituted (Y ) MeO, Me, H, Cl, and NO2) benzyl 3′-Z-substituted (Z ) CN, CONH2,
H, F, Ac) pyridinium chlorides in deuterium oxide. 3′-Cyanopyridine substrates reacted much faster
than nicotinamide and pyridine substrates; in the pyridine series the 4-Me, 4-H, and 4-Cl benzyl
analogs did not react for up to 6 months at 96 °C in 1.7 M NaN3. The 3′-cyanopyridine substrates
do not exhibit borderline kinetic behavior, but the nicotinamide substrates do. The Hammett plot
is flat for the NaN3 reaction of 3′-cyanopyridine substrates and increasingly V-shaped for the
nicotinamide and pyridine substrates. The values of âLG (four-point plot) for the NaN3 reaction of
the 4-MeO benzyl substrates is -1.45, which is usually interpreted as being a very “late” activated
complex. Two-point Brønsted “plots” for the other benzyl derivatives and for two N-methylpyri-
dinium ions give values of âLG in the same range. The second-order rate constant and activation
values for N-methyl-3′-cyanopyridinium iodide are within the same range as those for the benzyl
substrates. For the hydrolysis reaction, the Hammett plot is linear for 3′-cyanopyridine substrates
(F+ ) -1.24) and flat for the nicotinamide substrates. The extent of hydrolysis of 0.005-0.05 M
solutions of the 3′-cyanopyridinie substrates depended on the initial concentration of substrate,
and hydrolysis was slowed significantly or stopped completely in the presence of exogenous
3-cyanopyridine. These results show that an equilibrium is established among the products for
the 4-MeO, 4-Me, 4-H, and 4-Cl substrates; the 4-NO2 substrate reacted too slowly to discern any
difference. Data for the extent of hydrolysis were fitted by an equation derived assuming the
equilibrium. Despite this limitation on a classic test of mechanism, the rates and F values are
consistent with direct displacement by solvent and not with a unimolecular process. These results,
which are rationalized in terms of the Pross-Shaik model, suggest that there are no ion-dipole
complex intermediates in the benzyl series and show that borderline kinetic behavior is a function
of leaving group ability and is not necessarily related to a change in mechanism. A computational
approach was used to evaluate anomalous âLG values for the hydrolysis and nucleophilic substitution
reactions of the methypyridinium ion substrates. It was found that neither the Nu-substrate bond
lengths nor the difference in charge matched the âLG values. The value of ∆∆Sq of -15 gibbs/mol
between (4-methoxybenzyl)-3′-cyanopyridinium chloride and the corresponding dimethylsulfonium
chloride in the NaN3 reaction, which is the result of the solvation of the pyridine at the transition
state and the lack of solvation of SMe2, is used to argue that the source of NAD+ glycohydrolase
“catalysis” of NAD+ bond cleavage is the result of desolvation of the leaving group upon binding.
In tr od u ction
to exist as solvent-equilibrated intermediates or in
contact with a strong nucleophile such as N3- in solution.
Oppenheimer et al. have studied the cleavage of the
nicotinamide-ribosyl bond in nicotinamide-adenosine
dinucleotide (NAD+)1 and in a series of 2′-substituted
â-nicotinamide ribosides2 and arabinosides,3 and Schuber
and his colleagues have studied the solution and NAD+
glycohydrolase (EC 3.2.2.6)-catalyzed cleavage of NAD+
analogs with different pyridine leaving groups.4 All the
evidence suggests that bond cleavage is dissociative and
involves either an SN1 or an ion-dipole complex (desig-
nated IDC for solution reactions) mechanism. A purely
dissociative mechanism is at odds with a suggestion made
by J encks5 that sugar oxocarbenium ions are too unstable
In a computational study of the gas-phase dissociation
of â-nicotinamide riboside in the presence of several
water molecules, Schro¨der et al.6 found that the struc-
tures of the activated complexes and energies of the
transition states for the inversion and retention reactions
were different, which led them to question the general
validity of the “exploded” transition state invoked by
Sinnott and J encks7 to explain the relative reactivities
and selectivities of glucosyl pyridinium ions. A recent
study of the gas-phase dissociation of Handlon’s series2
of 2′-substituted â-nicotinamide arabinosides showed that
dissociation occurred through an ion-neutral complex
intermediate8 (designated an INC for the gas-phase
* Address correspondence to N.B. as follows: Phone: (707) 824-9770.
Fax: (415) 476-9687. E-mail: buckley@cgl.ucsf.edu.
X Abstract published in Advance ACS Abstracts, September 15, 1996.
(1) J ohnson, R. W.; Marschner, T. M.; Oppenheimer, N. J . J . Am.
Chem. Soc. 1988, 110, 2257-2263.
(2) Handlon, A. L.; Oppenheimer, N. J . J . Org. Chem. 1991, 56,
5009-5010.
(3) Handlon, A. L.; Xu, C.; Mu¨ller-Steffner, H.-M.; Schuber, F.;
Oppenheimer, N.J . J . Am. Chem. Soc. 1994, 116, 12087-12088.
(4) Tarnus, C.; Schuber, F. Bioorg. Chem. 1987, 15, 31-42. Tarnus,
C.; Muller, H. M.; Schuber, F. Ibid. 1988, 16, 38-51.
(5) Young, P. R.; J encks, W. P. J . Am. Chem. Soc. 1977, 99, 8238-
8248. Amyes, T. L.; J encks, W. P. J . Am. Chem. Soc. 1989, 111, 7888-
7900. Banait, N. S.; J encks, W. P. J . Am. Chem. Soc. 1991, 113, 7951-
7958.
(6) Schro¨der, S.; Buckley, N.; Oppenheimer, N. J .; Kollman, P. A.
J . Am. Chem. Soc. 1992, 114, 8232-8238.
(7) Sinnott, M. L.; J encks, W. P. J . Am. Chem. Soc. 1980, 102, 2026-
2032.
S0022-3263(96)00729-3 CCC: $12.00 © 1996 American Chemical Society

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 21, 1996 7361
Sch em e 1
reaction). Correlations among the measured relative
rates and the AM1 values of ∆Hq for the gas-phase
reaction and the experimentally determined values of
∆Gq for the solution reaction strongly suggested that the
mechanism of dissociation was the same in both phases,
but that the rate-limiting step was different.
It seemed that some insight into the question of
intermediate stability and mechanism might be gained
by studying the hydrolysis and nucleophilic substitution
reactions of pyridinium ion substrates with potentially
stable carbenium ion intermediates. Benzyl substrates
provide a convenient system because the structure of both
the putative intermediate and leaving group can be
varied easily. It was established with certainty recently
that the (4-methoxybenzyl)carbenium ion is an interme-
-
diate in the solvolysis and N3 substitution reactions of
neutral (benzoates and chlorides9 ) and charged (di-
methylsulfonium ion10,11 ) substrates. Other evidence12
suggests that (4-substituted benzyl)dimethylsulfonium
ions bearing substituents with σ+ > -0.31 react with
nucleophiles and water by direct displacement.
Katritzky13 has published a large amount of data on
the substitution reactions of benzyl and other substrates
with highly arylated or fused-ring pyridines as the LG
in chlorobenzene with neutral amine nucleophiles, pri-
marily piperidine. (4-Methoxybenzyl)-2′,4′,6′-triphenylpy-
ridinium ion and several of the fused-ring substrates
exhibit borderline kineticsskobsd does not double with a
doubling of the concentration of nucleophilesbut ben-
zyl triphenylpyridinium ion substrates bearing 4-Y-
groups with σ+ > -0.31 do not. The Hammett plot for
the second-order reactions of all triphenylpyridinium
ion substrates is linear (see below). Katritzky inter-
preted these results in terms of an IDC and not a mixed
mechanism (Scheme 1).
Ch a r t 1
ions are intrinsically more stable than the benzylcarbe-
nium ions in the gas phase and that the mechanisms of
hydrolysis and NaN3 substitution are different, with
unimolecular reactions for the ribosyl compounds and
bimolecular reactions for the benzylpyridinium ions.
(With our work on benzyldimethysulfonium ions, we have
a system with which to compare the effects of the LG.)
These results are inconsistent with the supposition that
the ribosyl oxocarbenium ions cannot exist as discrete
intermediates on the reaction coordinate for either solu-
tion or enzyme-catalyzed reactions.
We have prepared and studied the reaction with NaN3
in D2O of a series of (4-Y-benzyl)-3′-Z-substituted pyri-
dinium chlorides (Chart 1). Because the rate constants
could be measured conveniently, we have concentrated
on 1a -e. While most anionic nucleophiles add to the 2-
or 4-position of pyridinium ions to give “dead-end”
-
products that do not hydrolyze, N3 does not. Because
of this limitation on the range of nucleophiles, we can
obtain âLG but not âNuc. Unlike Katritzky’s results in
chlorobenzene with neutral nucleophiles, we found that
all substrates reacted with NaN3 in a strictly second-
order manner and that hydrolysis appears to occur by
direct solvent displacement.
With Maltby and Burlingame, we measured the rela-
tive rates for the gas-phase dissociation of these pyri-
dinium ion substrates and have reported the results
elsewhere.14 Thus, we have solution, gas-phase, and
computational results, obtained in the same laboratory
under essentially identical conditions, for the ribosyl and
benzylpyridinium ions that can be compared directly. In
brief, the comparison shows that the ribosyloxocarbenium
Exp er im en ta l Section
Gen er a l P r oced u r es. All chemicals were obtained from
Aldrich and used without further purification. NMR spectra
were recorded on a General Electric QE-300 FT-NMR fitted
with a variable-temperature probe rated at (0.1 °C. Liquid
secondary ion mass spectra (LSIMS) were obtained on a four-
sector Kratos Concept II in the positive ion mode in the UCSF
Mass Spectrometry Facility. Plots were made, and linear re-
gression was performed on Origin V2.8 or 4.0.
Syn th eses. All substrates were prepared by alkylation of
the appropriate pyridine with the appropriate benzyl chloride
using methods described elsewhere.14 Nicotinamide and 3-cy-
anopyridine were alkylated with MeI using the same proce-
dure. Appropriate 1H and 13C NMR and LSIMS spectra14 were
obtained for all compounds, which have been prepared previ-
ously as various salts (tosylates, triflates, iodides, etc.); they
were not further characterized.
(8) Buckley, N.; Handlon, A. L.; Maltby, D.; Burlingame, A. L.;
Oppenheimer, N. J . J . Org. Chem. 1994, 59, 3609-3615.
(9) Amyes, T. L.; Richard, J . P. J . Am. Chem. Soc. 1990, 112, 9507-
9512.
(10) Buckley, N.; Oppenheimer, N. J . J . Org. Chem. 1994, 59, 5717-
5723.
Kin etics. Two methods were used to obtain pseudo-first-
order rate constants (kobsd). For 2a -e and 3a -e, ca. 1-2 mM
solutions were made up in NMR tubes with the appropriate
NaN3 stock solution in D2O, with NaCl added to control ionic
strength, and heated in thermostated heat blocks (sand bath)
at 96 °C. The concentration of substrate was varied routinely
and had no effect on kobsd. Tubes were removed and spectra
(11) Kevill, D. N.; Ismail, N. H. J .; D’Souza, M. J . J . Org. Chem.
1994, 59, 6303-6312.
(12) Friedberger, M. P.; Thornton, E. R. J . Am. Chem. Soc. 1976,
98, 2861-2865.
(13) Reviewed in: Katritzky, A. R.; Brycki, B. E. Chem. Soc. Rev.
1990, 19, 83-105 and elsewhere.
(14) Buckley, N.; Maltby, D.; Burlingame, A. L.; Oppenheimer, N.
J . J . Org. Chem. 1996, 61, 2753-2762.

7362 J . Org. Chem., Vol. 61, No. 21, 1996
Buckley and Oppenheimer
Ta ble 1. Secon d -Or d er Ra te Con sta n ts for th e Rea ction
recorded at approximately 22 h intervals. For 1a -e, the
appropriate stock solutions of NaN3 in D2O were heated in the
probe of the NMR at various temperatures (40-90 °C), the
tube was removed, substrate was added to make ca. 5-10 mM
solutions, the tube was returned to the probe and shimmed
after ca. 1 min to allow equilibration, and spectra were
recorded at various time intervals. For particularly slow
reactionssthose that were complete within 4-5 hstubes were
held in thermostated heat blocks at the appropriate temper-
ature and transferred at various intervals to the heated NMR
probe in a round-robin fashion.
of 1a -e (80 °C) a n d 2a -e (96 °C) w ith Na N3 (k2) a n d
a
Wa ter (kw
)
Y
k2 (min-1 M-1
)
kw (min-1 M-1
)
krel
1a -e
MeO
9.6 ( 0.09 × 10-2
10.4 ( 0.4 × 10-2
8.3 ( 0.3 × 10-2
11.2 ( 0.5 × 10-2
9.1 ( 0.3 × 10-2
7.5 ( 0.2 × 10-2
23.9 × 10-6
7.8 × 10-6
3.1 × 10-6
1.8 × 10-6
0.3 × 10-6
5.9 × 10-5
0.4 × 104
1.3 × 104
2.6 × 104
6.2 × 104
36.0 × 104
0.14 × 104
Me
H
Cl
NO2
1-Me
The 2′ and 6′ pyridine protons rapidly exchanged for all
substrates under the reaction conditions. Depending on the
substrate, the benzyl methylene protons would also exchange,
slowest for the 4-MeO derivatives and fastest for the 4-NO2
derivatives. Therefore, rate constants were determined by
measuring the disappearance of the multiplet for the nonex-
changing pyridine 5′ proton. The total substrate was taken
as the sum of the substrate and product peaks, which provided
an internally consistent standard. As a check on the method,
in some instances the disappearance of the substrate benzyl
aromatic AA′BB′ peaks or the 4′ pyridine peak was monitored;
in all instances, the same rate constant was obtained.
The NaCl used to control ionic strength had no effect on
the rates. For instance, plots of kobsd vs [NaN3] for 1a -e
2a -e
MeO
Me
5.3 ( 0.7 × 10-4
5.3 ( 0.8 × 10-4
2.9 ( 0.1 × 10-4
2.4 ( 0.2 × 10-4
5.3 ( 0.8 × 10-4
2.3 ( 0.7 × 10-4
2.2 × 10-6
2.3 × 10-6
2.0 × 10-6
2.5 × 10-6
2.0 × 10-6
1.6 × 10-6
241
230
145
96
265
144
H
Cl
NO2
2-Me
a
All rate constants determined at µ ) 1.7 (NaCl). NaN3 range
0-0.85M for 1a , 2a -e, 2-Me; 0-1.7 M for 1b-e, 1-Me. For 1a -
e, all r g 0.999; for 2a -e, all r g 0.980. kw ) kobsd/55.5. For 1a -
e, kobsd for the water reaction was estimated from the plots in
Figure 6; for 2a -e and 2-Me, kobsd for the water reaction were
measured directly (as in Figure 2).
extrapolate to 0 at 80 °C; if there is a reaction with Cl-, it is
Ta ble 2. Eyr in g Activa tion Va lu es for th e Rea ction of
-
very slow and kCl would not contribute significantly to kobsd
.
a
1a -e a n d 1-Me w ith Na N3
For 1a -c, 2a -e, 1-Me, and 2-Me, hydrolysis at 96 °C in pure
D2O occurs faster than hydrolysis in 1.7M NaCl solutions,
consistent with an effect of ionic strength on rate and not with
a bimolecular reaction with Cl-.
Product analysis was performed by extracting reaction
mixtures with CDCl3 and integrating the appropriate aromatic
AA′BB′ peaks in 1H-NMR spectra.
Su p p r ession Stu d ies. For studies of the suppression of
hydrolysis in the absence of exogenous leaving group, serial
dilutions were made of stock solutions of the appropriate
pyridinium ion in D2O. Aliquots in NMR tubes were heated
in thermostated heat blocks (sand bath), and spectra were
recorded at intervals of 1-2 days. For studies of suppression
of hydrolysis with an exogenous leaving group for 1a -e, 5 mM
solutions of substrate containing 5 mM 3-cyanopyridine were
heated and analyzed as described above.
∆Hq
(kcal/mol)
∆Sq
(gibbs/mol)
∆Gq
80
Y
(kcal/mol)
r
MeO
Me
H
Cl
NO2
1-Me
17.6 ( 0.8
15.7 ( 0.6
14.4 ( 0.1
15.6 ( 0.4
14.8 ( 0.3
15.0 ( 0.1
-21.6 ( 2.0
-26.9 ( 1.6
-30.9 ( 0.4
-28.8 ( 1.3
-29.7 ( 0.8
-29.5 ( 0.2
25.2
25.3
25.3
25.2
25.3
25.8
0.9970
0.9980
0.9999
0.9992
0.9995
0.9999
a
Determined from four to five point plots for 1a -e and a three-
point plot for 1-Me.
Resu lts
Na N3 Kin etics. Values of pseudo-first-order kobsd were
obtained by linear regression as the slopes of plots of ln
(Ct/Co) vs time, where Ct was the relative concentration
at time t measured from the 5′-pyridine peak and Co was
the sum of the relative concentrations of Ct and product
pyridine. For 1a -e and 1-Me, plots were linear over
three to four measured half-lives and generally had r g
0.990; in several instances the r values were in the range
0.980 to 0.990. For 2a -e, 3a ,e, 5a ,e, and 2-Me, plots
were linear over 1-4 half-lives (see below). Values of
kobsd typically varied by 1-10% SE. Second-order rate
constants (k2) were determined by linear regression from
plots of kobsd vs [NaN3] (Table 1). In all but two instances,
r g 0.996. For 1a -e, plots were based on the averages
of two to eight determinations. For the much slower
reactions of 2a -e, 3a ,e, 5a ,e, and 2-Me, plots were based
F igu r e 1. Second-order rate plots for the reaction of 1a with
NaN3 (µ ) 1.7, NaCl) in D2O at 50 °C (b), 60 °C (9), 70 °C (2),
80 °C (1), and 90 °C ([).
80, and 90 °C (Figure 1). Curvature is most probably the
result of a negative primary salt effect expected for
reaction between two oppositely charged ions. These
plots appear to intercept the origin within error; as shown
below, however, there is a slow hydrolysis reaction.
Second-order rate constants for reaction at 80 °C (µ )
1.7, NaCl), determined in the range 0-1.7 M NaN3 (all r
> 0.999), are listed in Table 1.
Second-order rate constants for 1-Me were determined
at 70, 80, and 90 °C from plots of kobsd vs [NaN3] (0-1.7
M, r > 0.996). Unlike the benzyl substrates, however,
there were non-zero intercepts for the second-order plots,
with rate constants in the range 1.5 × 10-3 to 17 × 10-3
min-1 for the three temperatures. An extrapolation gives
on duplicate determinations of kobsd
.
Activation values for 1a -e were determined from four-
to five-point Eyring plots (range 50-90 °C) by linear
regression (r g 0.998). For 1-Me iodide, only three points
were used (70-90 °C, r ) 0.9999). Values summarized
in Table 2 are reported ( SD.
(a ) 3-Cya n op yr id in es. These substrates reacted
smoothly with NaN3 at convenient rates in the range 50-
90 °C. Plots of kobsd vs [NaN3] were linear over more than
half of the concentration of NaN3, after which a slight
curvature could be seen, especially for the plots at 70,

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 21, 1996 7363
F igu r e 2. Rate plots for the reaction of 2a with NaN3 (µ )
1.7, NaCl) in D2O at 96 °C. [NaN3] ) 1.7 (b), 0.85 (9), 0.43
(2); O ) water only. The r values are 0.9994, 0.9995, 0.9962,
and 0.9992, respectively.
F igu r e 4. Reaction progress for 1a ,b at 80 °C in pure D2O as
a function of the initial molar concentration of substrate: 0.005
) ×; 0.010 ) [; 0.020 ) 1; 0.030 ) 2; 0.040 ) 9; 0.050 ) b.
the slower reactions of 1c-e at 96 °C (5, 20, and 50 mM
solutions; not shown). This curvature is the same as
found for (4-methoxybenzyl)dimethylsulfonium chloride
that we have shown10 is caused by the establishment of
the equilibrium RX+ h ROH + H3O+ + X° during the
course of the reaction. For the sulfonium ions, data were
fitted to a cubic equation derived for the equilibrium that
assumed X° is not protonated, which is a reasonable
assumption for SMe2 (pKa ) -6). For pyridinium ions,
however, a significant proportion of the liberated 3-cy-
anopyridine (pKa ) 1.45) would be protonated. With the
simplifying assumption that [ROH] ∼ [XH+], the equation
for the equilibrium is
F igu r e 3. Second-order rate plot for the reaction of 2a with
NaN3 (µ ) 1.7, NaCl) in D2O at 96 °C. The dashed line is a
regression fit (r ) 0.990) to the point for water alone and 0.43
M and 0.85 M NaN3.
a value of ca. 40 × 10-3 min-1, or kw ) 7.2 × 10-4 M-1
min-1, for the second-order rate constant at 96 °C.
(b) Nicotin a m id es. Rate plots for reaction of NaN3
with 2a -e were linear over 2-4 half-lives at 96 °C
(Figure 2). Plots of kobsd vs [NaN3] are linear between 0
and 0.85 M NaN3, after which curvature appeared in
some plots (Figure 3). Curvature could be the result of
the negative primary salt effect or it could be the result
of NaN3-catalyzed hydrolysis of the amide; because the
pKa for nicotinic acid is higher than that for nicotinamide,
the displacement reactions would be slower and could
account for the curvature. These plots had non-zero
intercepts (khyd), and there was a measurable reaction
in water alone; the rate constants were the same within
error. Values for the second-order rate constants were
taken from the linear portions of the curves (three
points); these and the values for kw, which for comparison
with the NaN3 numbers are the second-order rate con-
stants khyd/55.5, are listed in Table 1.
(c) P yr id in es. 3a and 3e reacted very slowly with
NaN3 at 96 °C. Second-order rate constants, based on
1-2 half-life kobsd values, are listed in Table 1. For 3b-
d , however, there was no detectable reaction at 96 °C in
up to 1.7 M NaN3 for 6 months!
D2O On ly Kin etics. (a ) 3′-Cya n op yr id in es. In D2O
alone, the extent of reaction is a function of the concen-
tration of substrate. Shown in Figure 4 are plots of ln
(Co/Ct) vs time for the hydrolysis of 5-50 mM solutions
of 1a and 1b at 80 °C; a similar pattern was found for
Co ) KaKeqf/(1 - f)2
where Co is the initial concentration of substrate, Ka is
the acidity constant for the pyridine, Keq is the equilib-
rium constant, and f is the fraction remaining at equi-
librium ([RX+]200h/Co). The fit of data for 1a to this
equation (Co ) 5-50 mM at t ) 200 h) by the Marquand-
Leverberger nonlinear least-squares algorithm in the
Origin software is shown in Figure 5 (Keq ) 2.29 ( 0.06,
ø2 ) 3.84 × 10-6).
Studies of exogenous LG were limited for sulfonium
ions by the solubility of SMe2 in water.10 Solubility of
the pyridine LGs is not a problem, but it is difficult to
analyze and quantitate the aromatic region of spectra
containing large amounts of exogenous pyridine; for 1a
and 1b, however, the methyl signals can be followed
conveniently, and the same is true for the methylene
signals for 1c and 1d , although for high concentrations
of pyridines some exchange of these protons takes place
and the readings can be inaccurate. Shown in Figure 6
are plots of ln (Co/Ct) for the hydrolysis of 5mM solutions
of 1a -d in the presence of 5 mM 3-cyanopyridine (closed
circles) at 80 °C; for comparison, the plots for 5 mM sub-
strate alone are included (open circles). The presence of
exogenous leaving group further suppresses the reaction.

7364 J . Org. Chem., Vol. 61, No. 21, 1996
Buckley and Oppenheimer
F igu r e 5. Fit of the data for 1a from Figure 4 to the equation
F igu r e 7. Rate plots for hydrolysis of 2a (almost certainly
hydrolyzed to the 3′-carboxy compound 4a ) in 5.3 M DCl at
80 °C (b), conditions that would prevent the equilibrium among
starting material and products. The plot is linear (r ) 0.999)
over >3 half-lives. For reference, the hydrolysis of 0.005 M
4a at 80 °C in pure D2O is shown (O).
for the equilibrium RPy+ / ROH + Py + H3O+ D PyH+. Keq
)
2.29 ( 0.06, ø2 ) 3.8 × 10-6
.
under the reaction conditions, and the pyridinium ion and
aromatic protons cannot be resolved well from those for
product pyridine or alcohol. Nonetheless, spectra for 5,
30, and 50 mM solutions and 5 mM substrate plus 5 mM
3-cyanopyridine change only very slowly and to the same
extent (measured by noting the change in the relative
heights of the substrate and product 5′-proton, which is
difficult to quantitate with any accuracy). We did not
attempt to characterize this reaction further.
(b) Nicotin a m id e. The same concentration depen-
dence found for the 3′-cyanopyridine substrates was seen
for the nicotinamide substrates at 96 °C; the reaction was
slower, and the extent of reaction was less than for 1a -e
(not shown). We did not repeat the reaction in the
presence of nicotinamide.
In the (4-methoxybenzyl)dimethysulfonium ion work,
we showed that if SMe2 were removed as the Hg+2 or Zn+2
complexes, the reaction went to completion with no
suppression.10 In 5.3 M DCl at 96 °C, the amide moiety
of nicotinamide would be hydrolyzed rapidly to the acid,
and all nicotinic acid released by solvolysis would be
protonated and the equilibrium with starting material
would be suppressed; it is also probable that the bis
benzyl ether, which would have essentially the same
NMR spectrum as the alcohol, would form under these
extremely acidic conditions, which would further sup-
press the equilibrium.12 In the event, solvolysis of 2a in
5.3 M DCl at 96 °C proceeded smoothly, and the plot of
ln (Ct/Co) vs time was linear over 3 half-lives (Figure 7;
r ) 0.999). Similar plots were obtained for hydrolysis in
the presence of 2 M D2SO4 (not shown). For comparison,
the rate plot for the same concentration of 4a , which was
prepared by saponification of the methyl ester, in D2O
(µ ) 1.7, NaCl) is shown in Figure 7; the estimated rate
constant in DCl is 35-fold higher than in D2O alone, and
the plot for D2O alone is slightly curved.
F igu r e 6. Reaction progress for 0.005M 1a -d at 80 °C in pure
D2O with 0.005 M 3-cyanopyridine added (b) and in pure
D2O (O).
(c) Oth er Der iva tives. The same concentration-
dependent pattern was found for hydrolysis of 5a and
6a (plots not shown). Thus, the equilibrium among
starting material and products is independent of the pKa
of the leaving group and is general for the class.
Lin ea r F r ee En er gy Rela tion s (LF ERs). (a ) Ha m -
m et t P lot s. Hammett plots for the NaN3 reaction of
1a -e (80 °C), 2a -e (96 °C), and 3a ,e (96 °C) are shown
in Figure 8. For 1a -e, the plot is flat. (Hammett plots
based on k2 values for lower temperatures show small
increases to a V-shaped curve with decreasing temper-
Because of the severe curvature and the fact that use
of either the Guggenheim method or inclusion of an
infinity value in the rate equation failed to give linear
plots, it was difficult to obtain accurate values for the
hydrolysis rate constants kw. Points taken during the
first half-life give approximately linear rate plots, how-
ever, and the values of kw can be estimated. These are
listed in Table 1.
NMR spectra for 1e are very difficult to analyze
because the benzyl methylene protons exchange rapidly

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 21, 1996 7365
F igu r e 10. Brønsted plot for the NaN3 reaction of (4-
methoxybenzyl)pyridinium ions at 80 °C (µ ) 1.7, NaCl) (b);
âLG ) -1.45 (r ) 0.997). Data for the reactions of N-
methylpyridinium ions 1-Me and 2-Me with NaN3 under
identical conditions are shown (]); âLG ) -1.56.
F igu r e 8. Hammett plots for the reaction with NaN3 (k2) of
1a -e (b), 2a -e (9), and 3a ,e (2) (µ ) 1.7, NaCl). The point
for 3c was estimated as described in the text.
our assignment of mechanism is of necessity circumstan-
tial. The Hammett plot for hydrolysis of 2a -e (Figure
9), based on directly measured rate constants, is es-
sentially flat.
(b) Br øn sted P lots. Values of âLG at 96 °C were
determined from plots of log k2 vs pKa of the pyridine for
1-3a and 5a and for 1-Me and 2-Me. Values of k2 used
in these plots were based on at least duplicate determi-
nations; the second-order rate constants at 96 °C for the
1a and 1-Me were obtained from extrapolation of the
Eyring plots. The Brønsted plot for 1-3a and 5a (Figure
10) is linear with âLG ) -1.47. The two-point “plot” for
1-Me and 2-Me (Figure 10) has a slope of -1.56; two-
point “plots” for 2b-d and 3b-d (not shown) are parallel
to the four-point line for 1-3a and 5a . (The Hammett
plot (vs σm) for the effect of substituents on the pyridine
in 1a is linear [not shown; r ) 0.991] with F ) 8.9.)
For the hydrolysis reactions, rate constants determined
for 1a -c at 96 °C in single experiments were within
factors of 2-4 of the rate constants for 2a -c, and were
subject to the uncertainties noted above. These give
approximate âLG values in the range -0.15 to -0.30.
For 1-Me and 2-Me, however, the value of âLG calculated
from the rate constants based on the intercept values for
1-Me NaN3 second-order plots and the directly measured
value for 2-Me is -1.4, on the same order as the NaN3
value.
F igu r e 9. Hammett plots for the hydrolysis in D2O of 1a -e
(b) and 2a -e (0) at 80 and 96 °C respectively (µ ) 1.7, NaCl).
For 1a -e, F+ ) -1.24 (r ) 0.997); for 2a -e, F+ ) 0.05 (r )
0.26). For reference, the Hammett plot is shown for Katritzky’s
second-order rate constants13 for the reaction of benzyl tri-
phenylpyridinium ions with piperidine in chlorobenzene at 100
°C (×); F+ ) -0.51 (r ) 0.960).
ature. The change is not great, however). For 2a -e, the
plot is V-shaped with the break at 2c. As noted above,
3b-d did not react with NaN3 after 6 months at 96 °C;
if it is assumed, however, that 3% reactionsthe limit of
NMR detectionshad occurred for 3c, the three-point plot
in Figure 8 is obtained. This estimated point is the upper
limit for k2; it is entirely possible that the “real” value
would be lower, and the break would be deeper. The
salient feature is that with increasing basicity of the LG,
the depth of the break in the Hammett plots increases.
The Hammett plot for the hydrolysis of 1a -e (Figure
9), which is based on estimated rate constants, is linear,
which indicates the same mechanism of hydrolysis for
all substrates. For (4-methoxybenzyl)dimethylsulfonium
chloride, the rates of hydrolysis and second-order reaction
with NaN3 are within the same range, and we have
shown10 that there is a change in mechanism from SN1
for the (4-methoxybenzyl)dimethylsulfonium compound
to SN2 for substrates with less electron-donating 4-sub-
stituents, which react by direct displacement by solvent.
Because of the great difference between the hydrolysis
and NaN3 second-order rate constants for 1a -e, we as-
sume that hydrolysis takes place by direct solvent dis-
placement. The standard test for this, common leaving
group suppression, cannot be applied here because of the
equilibrium among products and starting materials, so
P r od u ct Stu d ies. The only product detected in NMR
spectra of CDCl3 extracts of reaction mixtures of 1a -e
was the corresponding benzyl azide. For 2a -e, ∼5-10%
of the respective alcohol could be detected, which is
consistent with the concomitant hydrolysis reaction;
the majority of the product was the respective benzyl
azide. No alcohol was found for 3a and 3e, but because
of the slow rate, these reactions were not taken to
completion.
Discu ssion
Borderline kinetics in the substitution reactions of
benzyldimethylsulfonium ions with NaN3 in water15 and
for reactions of benzylpyridinium ions with neutral
amines in chlorobenzene13 have been interpreted in terms
of an IDC mechanism; in both cases a mixed SN1/SN2
mechanism was rejected for various reasons (Scheme 1).
(15) Sneen, R. A.; Felt, G. R.; Dickason, W. C. J . Am. Chem. Soc.
1973, 95, 638-639.

7366 J . Org. Chem., Vol. 61, No. 21, 1996
Buckley and Oppenheimer
Borderline kinetics in the reactions of N-(methoxy-
methyl)-N,N-dimethylanilinium ions with nucleophiles,
however, was interpreted by Knier and J encks16 as
evidence for a concerted mechanism that is “enforced”
because the putative cationic intermediate in an IDC or
stepwise mechanism, MeO)CH2+, was judged to be too
unstable to exist as a distinct, solvent-equilibrated spe-
cies. Using semiempirical methods we have recently
examined the gas-phase dissociation of a number of
compounds with a variety of oxocarbenium ion interme-
diates that can form by dissociation of different LGs
(NMe2Ar, Py, H2O, MeOH, Me2O)17 and found that
neither MeOCH2NMe2Ar+ nor MeOCH2Py+ had a distinct
transition state in either PM3 or AM1, consistent with
the arguments of Knier and J encks.16 Sulfur-containing
linear primary systems such as MeSCH2NMe2Ar+ and
secondary linear systems such as MeOCHMeNMe2Ar+-
and MeSCHMeNMe2Ar+ and the respective pyridinium
ions, protonated hemiacetals, methyl acetals, and di-
methyloxonium ions gave distinct transtion states, how-
ever, with the INCs much more stable for the R-sulfur
than for the R-oxygen compounds.17 These results show
an LG and substituent dependence that is not consistent
with the general view on oxocarbenium ion stability.18
An often unstated assumption is that a substitution
reaction must occur through a single mechanism; in
J encks’s reviews on these matters,18 the case is always
cast in terms of either a stepwise or concerted mechanism.
The last sentence of the Knier-J encks paper16 states the
consequences of this assumption most boldly: “[A]ll
solvolysis and substitution reactions at saturated carbon
that proceed through SN2 displacement mechanisms do
so simply because the intermediate in the alternative SN1
mechanism is too unstable to exist [italics added].” It
seems reasonable, however, that if ∆Gq of competing uni-
and bimolecular mechanisms are of the same magnitude,
there is no reason why simultaneous stepwise and con-
certed reactions with nucleophiles cannot occur, as found
recently for neutral9 and charged10,11 4-methoxybenzyl
substrates. It seems equally true that there is no reason
to believe that an SN1 reaction must occur for a substrate
that contains an excellent incipient carbenium ion (as in
1-5a ). Furthermore, borderline kinetics do not neces-
sarily have to signal the presence of borderline mecha-
nisms involving solvolytically generated ion-pair inter-
mediates or some intermediate (as opposed to direct)
solvent nucleophilic participation.
of solvent on the intrinsically unstable IDC, but no
alcohol products were detected for any 3′-cyanopyridine
substrate.
For the nicotinamide substrates 2a -e, there is bor-
derline behavior at 96 °C. The rates of the NaN3 and
hydrolysis reactions for 2a -e are comparable, as shown
by the rate plots for 2a in Figure 2; the ratios of the
second-order rate constants kAz/kw are between 96 and
265, which is relatively constant and much lower than
the range found for 1a -e. Alcohol products are also
found in the NaN3 reaction of the nicotinamide sub-
strates. This pattern of reactivity suggests that in these
systems the switch from no borderline to borderline
behavior is related to factors other than a change in
mechanism. The analysis presented below suggests that
in this system, at least, borderline kinetics is related to
the LG ability and to the different effects of charged or
neutral nucleophiles on the structure of the activated
complex for an uncomplicated concerted displacement
reaction.
LF ER Cor r ela tion s. The Hammett plots for the
NaN3 reaction show that the effect of the 4-substituents
on the rate constants depends on the LG. The plot for
1a -e is essentially flat, and the plot for 2a -e is slightly
V-shaped; the value of F for 2a -d is ca. -0.7 (Figure 8).
For the pyridine series, with the point for 3c based on
an estimated upper limit for the rate constant (see
Results), the Hammett plot is even more V-shaped. The
value of âLG for the 4-methoxybenzyl substrates is -1.45
(Figure 10), and the Hammett plot based on σm values
for the pyridine 3′-substituents is also quite large (+8.9,
r ) 0.991, not shown). Values for the other substituted
benzyl substrates, based on more limited rate data, are
in the same range. Therefore, the activated complex for
the NaN3 reaction is very “late” for all substituents.
Westaway and Ali19 have shown that a change from a
better to a worse leaving group (for the reaction of
thiophenoxides with substituted N-benzyl-N,N-dimeth-
ylanilinium ions in DMF, a system in which the LGs have
the same pKa range as the pyridines used here) “loosens”
the Nu-benzyl methylene bond with little effect on the
benzyl methylene-LG bond. Kurz and El-Nasr20 found
the same effect in k14/k15 ratios for the reactions of various
pyridines with neutral MeX derivatives (X ) I-, -OTs,
-
and OTf). This “anti-Hammond” effect is present in our
system as well. For the NaN3 reaction, the tightness of
the Nu-benzyl methylene bond in the activated complex
would be expected to decrease as a function of the LG in
the order 1 > 2 > 3. The extent of substituent interaction
with the reaction center would be expressed in the
same order: The more fully formed the Nu-benzyl
methylene bond, the less charge at the reaction center
and the less conjugation to the substituents, which is the
trend seen in the Hammett plots. While this is not a
direct resonance interaction as found in R-MeO- or
R-N3- substituted benzyl systems, the decreased reso-
nance interaction of 4-substituents on the phenyl ring
with the reaction center is the same as that described
recently by Richard et al.21 Thus, stabilization of the
activated complex is related more to the nucleophile-
carbenium ion interaction, which is mediated by the pKa
of the LG, than to the relatively constant LG-carbenium
ion interaction.
Our results show that borderline kinetic behavior is
variable across a series of substrates in which only minor
structural changes are made in the LG at some distance
from the reaction center. There is no borderline behavior
in the reaction of 1a -e with NaN3 in water at 80 °C
(Figure 1 for 1a ). A slow reaction with water can be
measured for 1a -e, however. The relative rates for the
second-order reactions kAz/kw, where kw is kobsd for the
water reaction/55.5, range from 4 × 103 for 1a to 3.6 ×
105 for 1e (Table 1). While we cannot rule out an SN1
reaction with absolute certainty because of the equilib-
rium among products and starting material (see Results),
the much slower hydrolysis reaction is consistent with
solvent displacement on 1a -e. Moreover, if an IDC were
an intermediate in the pyridinium ion reactions, it would
be expected that an alcohol would be formed by attack
(16) Knier, B. L.; J encks, W. P. J . Am. Chem. Soc. 1980, 102, 6789-
6796.
(17) Buckley, N.; Oppenheimer, N. J . J . Org. Chem., in press.
(18) J encks, W. P. Acc. Chem. Res. 1980, 13, 161-169. J encks, W.
P. Chem. Soc. Rev. 1981, 10, 345-375.
(19) Westaway, K. C.; Ali, S. F. Can. J . Chem. 1979, 57, 1354-1367.
(20) Kurz, J . L.; El-Nasr, M. M. J . Am. Chem. Soc. 1982, 104, 5823-
5824.
(21) Richard, J . P.; Amyes, T. L.; J agannadham, V.; Lee, Y.-G.; Rice,
D. J . J . Am. Chem. Soc. 1995, 117, 5198-5205.

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 21, 1996 7367
This qualitative description of the effects of nucleo-
phile, LG, and substituents on the activated complex is
in complete accord with the Pross-Shaik22,23 avoided
crossing model. Of the four canonical configurations for
the NaN3 reaction, the form corresponding to product,
Nu‚‚R :X°, is clearly the more important for a tight
Nu-benzyl bond (as in 1a -e). As this bond is “loosened”
with a change to a worse LG, the forms Nu:-R+:X° and
Nu‚ R:- ‚X° become more important; the effects of elec-
tron-donating substituents would be expressed in stabi-
lization of R+ in the former and would decrease in the
normal manner, while an electron-withdrawing group
would stabilize the latter, and accordingly the rate of the
4-NO2 substrate is increased relative to H (as in 2a -e
and 3a ,e). There is also a direct LG effect that is in-
dependent of the effect on the nucleophile interaction
with the reaction center. Addition of electron-withdraw-
ing substituents to the pyridine improves the LG ability
by lowering the pKa and increases the ionization potential
of the LG, which makes Nu‚ R:- ‚X° less important for
transition state mixing. The “tightening” effect on the
nucleophile-carbenium ion center for a better LG and the
effect of the LG on the stabilization of the canonical forms
lead to a flattening of the Hammett plot for 1a -e (Figure
8). Thus, our results are in complete accord with Pross’s
rationale for the shape of Hammett plots for benzyl
substrates.23
substrates vs a three (or more)-form mix for substrates
that can stabilize the activated complex by resonance (as
benzyl and MeOCH2- substrates may do), this does not
appear to be the entiresor even correctsstory.
Sufficient data are not available for the benzylpyri-
dinium ions to permit analysis. There are enough data,
however, from the work reported here and by Arnett and
Reich24 and Metzger’s group25 for MePy+, and from Knier
and J encks16 for MeOCH2-NMe2Ar+ substrates, to ana-
lyze the effect of the LG on reaction with a range of
negative and neutral nucleophiles, including water. The
stability of the activated complex is a function of the
interaction of both the LG and the Nu with the reaction
center, and stability should be reflected in some measure
of the rate.
â
LG’s for the Me substrates cover a wide range (-0.4
to -1.5) and appear to be independent of the charge on
the nucleophile. Values for the MeOCH2- substrates16
are in a much narrower range (-0.54 to -0.9) and also
appear to be independent of the charge on the nucleo-
phile. The interaction of the Nu and LG with the reaction
site can be evaluated with cross-correlation plots26 of âLG
vs the Swain-Scott nMeI value for the reactions of
MePy’s+ with N3-, I-, Ph3P, and water,27 and for MeOCH2-
NMe2Ar+ with HO-, AcO-, three primary amines, and
water.16 While these plots are generally used to evaluate
the effect of a change in a substituent on a general acid
or on a substrate, in principle there is no reason why they
cannot be used to evaluate the equivalent effect for
changes in the reactivity of nucleophiles and LGs.
For the various nucleophiles these plots show, in each
case, a good linear relation for all nucleophiles except
water, the points for which are significantly off the
correlation lines (Figure 11). The slope of these plots is
a measure of the sensitivity of the substrate to the
nucleophile and is smaller for the MeOCH2-substrate
(0.13) than for the Me (0.43), or “earlier” for the stabilized
substrate. Even though it might be argued that the point
for MeOCH2NMe2Ar+ is off the line because the hy-
drolysis of this substrate is stepwise with the formation
of an oxocarbenium ion too unstable to exist outside the
initial solvation shellsan argument made recently by
Banait and J encks28 for the fate of the glycosyl oxocar-
benium ion in the hydrolysis of R-fluoroglucosesand thus
represents a change in mechanism, this is certainly not
true for the methyl compounds. The data for both
substrates are quite consistent with direct solvent dis-
placement.
The Hammett plots for the hydrolysis reaction of 1a -e
and 2a -e show trends opposite those found for the NaN3
reaction (Figure 9). The linearity of the plots for 1a -e
and 2a -e indicates a single mechanism, and our as-
sumption of reaction through solvent displacement is
consistent with the F value of -1.24 for 1a -e. The
Hammett plot for the hydrolysis of 2a -e is essentially
flat. Thus, while there is no indication of a change in
mechanism, it is clear that substituent effects as a
function of the LG are expressed differently for a neutral
than for a negative nucleophile. While we did not
measure the rate constants for the hydrolysis of all benzyl
substrates at 96 °C, second-order rate constants for 1c,d
(ca. 6-10 × 10-6 M-1 min-1) are within factors of 3-5 of
the second-order rate constants for 2c,d , which suggests
that the value of âLG for the hydrolysis reaction would
be low (ca. -0.15 to -0.30 based on the approximate
relative rates.)
For the reaction of a neutral nucleophile, the most
important canonical form for the activated complex is
Nu:° R+ :X°. Because the neutral nucleophile would not
interact with the reaction center as strongly as a nega-
tively charged species, the bond length should remain
relatively constant with a change in LG. The more basic
LG would stabilize R+, and in a “tight” activated complex
for both the nucleophile and LG this interaction may
lessen or even eliminate the effects of substituents,
which is the trend found: the Hammett plot is flat for
the more basic pyridine LG (Figure 9), and the âLG is low
(“tight”).
It seemed that an answer to this question could be
obtained using computational methods. Complete energy
surfaces were calculated in AM129 for the reactions
X°/- + R-3-CNPy+ h XMe+/o + 3-CNPy
for R ) Me (1Me) and R ) MeOCH2- with the same
(24) Arnett, E. M.; Reich, R. J . Am. Chem. Soc. 1980, 102, 5892-
5902.
(25) Berg, U.; Gallo, R.; Metzger, J . J . Org. Chem. 1976, 41, 2621-
2624.
(26) J encks, D. A.; J encks, W. P. J . Am. Chem. Soc. 1977, 99, 7948-
7960.
(27) Data for MePy’s were obtained in different solvents at different
temperatures; while these conditions affect the absolute rates, they
should not affect the Brønsted coefficients. As shown in Figure 9 of
ref 24, plots of log k vs pKa for the Menschutkin reaction of pyridines
with a variety of electrophiles in solvents that span the range from
DMSO to water to nitrobenzene and nitroalkanes over a rate range of
109 (!) all have essentially the same slopes. Microscopic reversibility
requires that the same is true for the reverse reaction.
(28) Banait, N. S.; J encks, W. P. J . Am. Chem. Soc. 1991, 113, 7951-
7958.
There is a caveat, however. The âLG values for the
NaN3 reaction for methyl and benzyl substrates are
essentially the same (Figure 10), but the trend for the
hydrolysis reaction is reversed, with âLG -0.3 for the
benzyls and âLG ) -1.4 for 1Me and 2Me. While results
such as these may be rationalized, as Pross has sug-
gested,22 by differences between a two-form mix for Me
(22) Pross, A. Adv. Phys. Org. Chem. 1985, 21, 99-196.
(23) Pross, A.; Shaik, S. S. Acc. Chem. Res. 1983, 16, 363-370.

7368 J . Org. Chem., Vol. 61, No. 21, 1996
Buckley and Oppenheimer
F igu r e 11. Cross-correlation plots for the reactions of MePy’s
with NaN3 (this study), I- (ref 27), Ph3P (ref 28), and water
(this study) and for the reactions of MeOCH2NMe2Ar+ with
HO-, AcO-, three primary amines, and water (ref 16). For
MePy, r ) 0.986; for MeOCH2NMe2Ar+, r ) 0.997.
F igu r e 12. Contour plots of the AM1 reaction surface for the
reaction of 1Me and 1MeO with water. The reaction begins
in the lower right-hand corner. The relative energies (in kcal/
mol) are given on the contour lines. The C-Py bond lengths
are the same, but the C-O bond lengths are very different.
Note that the surface for the MeOCH2- substrate is very
similar to the PM3 surface calculated for the inversion water
reaction of â-nicotinamide riboside (ref 6).
3-cyanopyridine LG30 (1MeO). The nucleophiles studied
were water (Figure 12 for the contour plots; the full
surface is shown in Figure S1, supporting information),
N3-, I- (Figure 13 for the contour plots; the full surface
is shown in Figure S2, supporting information), and
Me3P, the latter used rather than Ph3P for computational
simplicity. While a full report will be published else-
where with data for the benzyl systems and dimethyl-
aniline LGs, the results of the initial part of the study
suggest an answer to this apparent dilemma.
The computed ∆Hq and ∆HR values for the reaction of
1Me and 1MeO with all nucleophiles correlate very well
(Figure 14). The slope is smaller for 1MeO than for 1Me,
consistent with stabilization of the activated complex by
resonance.31 In terms of the energies involved, these are
perfectly well-behaved systems that follow the rate-
equilibrium criterion of the Hammond-Leffler postu-
late.32
Pross22 has argued that the Brønsted parameter has
no relation to charge development for MeX substrates,
which have only reactant and product canonical forms
that can mix to give the activated complex, but may be a
relative measure of charge development in stabilized
systems such as MeOCH2X, in which a third canonical
+
form involving MeOdCH2 can mix into the activated
(30) The pyridine LG was used for consistency and computational
simplicity. The pKa’s of 3-CN-pyridine (1.45) and N,N-dimethyl-3-
nitroaniline (2.65) are in the same range, and ∆HR values and HOMO-
LUMO gaps for the model reactions, which are thermodynamic
measures of reactivity, for both LGs correlate excellently (r > 0.999 99)
between the series. The plots of ∆HR vs ELUMO - EHOMO for the aniline
LG for either the Me or MeOCH2 substrate are the same as found for
the pyridine LG (Figures 17 and 18). (These plots are given as
supporting information.) These correlations show that the pyridine can
be used for direct comparison with the experimental data. The rates
for the reactions will be different because of steric and solvation factors,
however.
(31) This effect is opposite that predicted by Dewar and Dougherty
(Dewar, M. J . S.; Dougherty, R. C. The PMO Theory of Organic
Chemistry; Plenum: New York, 1975; pp 262-263) on the basis of an
analysis of the orbital interactions of the Nu and MeO MOs in the
activated complex. The solution data are in agreement with our
calculations; extrapolation of the Knier-J encks data for water reaction
of the m-nitrodimethylaniline substrate shows that it reacts 4800-fold
faster than the unstabilized 1Me at 80 °C.
(29) Reaction surfaces were computed in AM1 (Dewar, M. J . S.;
Zoebisch, E. G.; Healy, E. F.; Stewart, J . J . P. J . Am. Chem. Soc. 1985,
107, 3902-3909) using the reaction coordinate method. Enthalpies of
formation (∆Hf) for various structures were computed by varying either
the Nu or LG bond length with the other held constant; for the negative
Nu, the reaction coordinate was constrained to a torsional angle of
180°. The relative enthalpies for bond cleavage were obtained by
subtracting the ∆Hf of the initial complex formed between substrate
and Nu. After the TS had been located approximately on the surface,
intermediate structures were refined until a structure was obtained
that gave a single negative first normal mode on vibrational analysis
(McIver, J . W., J r.; Komornicki, A.. Chem. Phys. Lett. 1971, 10, 303-
306). Enthalpies of reaction (∆HR) were calculated from the ∆Hf of the
separated reactants and products. ∆HR computed in PM3 (Stewart, J .
J . P. J . Comput. Chem. 1989, 10, 209-220; 221-264) had different
values for each structure, but the trends found in AM1 were the same.
The values of EHOMO and ELUMO in AM1 and PM3 were very similar,
however, and showed the same trends.
(32) (a) Hammond, G. S. J . Am. Chem. Soc. 1955, 77, 334-338. (b)
Leffler, J . E. The Reactive Intermediates of Organic Chemistry;
Interscience: New York, 1956.

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 21, 1996 7369
F igu r e 14. Rate-equilibrium AM1 enthalpy correlations for
1Me (b) and 1MeO (9); the slopes are 0.21 (r ) 0.999) and
0.10 (r ) 0.999), respectively. Left to right the points are for
I-, N3-, Me3P, and water.
F igu r e 15. Charge density plots for the reaction of 1MeO
with water and Me3P. There is a large electrostatic repulsion
between the incoming water and the MeO that is not present
in the reaction with Me3P.
F igu r e 13. Contour plots of the AM1 reaction surface for the
reaction of 1Me and 1MeO with I-. The reaction begins in
the lower right-hand corner. The relative energies (in kcal/
mol) are given on the contour lines. The C-Py bond lengths
are essentially the same, and the “early” C-I bond lengths
have a bond order of ∼0.3. The same general shape and pattern
was found for N3- and Me3P. Note that while the second-order
for the water reactions. In neither system do the bond
lengths to the Nu or LG correlate with the âLG
.
The difference between the O-C TS bond lengths for
1Me and 1MeO is probably the result of charge repulsion
in the activated complex for 1MeO. As the activated
complex is formed in both substrates, a net positive
charge builds up on the water oxygen. A repulsive
interaction with the positively charged pyridine nitrogen
would be partially shielded by the methyl in both. In
1Me, there is no other interaction; in 1MeO, however,
as the resonance interaction develops and positive charge
is building up on the MeO oxygen, electrostatic repulsion
between the two oxygens would cause an earlier TS
(Figure 15). The greater stabilization afforded by the
resonance interaction is sufficient to overcome the energy
of this repulsive term. This interaction is avoided in the
phosphine TS because of either or both the greater
polarizability of P or because of hyperconjugation that
places the bulk of the charge on the phosphine methyls;
accordingly, the reaction surface for this neutral Nu
closely resembles that for the negative Nu’s.
The reason for the longer length of the C-N bond in
the water reaction is apparent from an examination of
the frontier MOs for the different Nu’s. Upon formation
of the initial complex, the HOMO of the entire complex
remains localized on the Nu for all complexes except the
two water structures, in which the HOMO is localized
on the LG (Figure 16). The LUMO of the initial complex
is localized on the water. Both the HOMO-LUMO gap
(ELUMO for the substrate - EHOMO for the Nu, Figure 17)
and the EHOMO for the activated complex (Figure 18) show
the same relationship found in the cross-correlation plots
(Figure 11). These FMO localizations are the result of
-
rate constants for the N3 and I- reaction of the (methoxy-
methyl)dimethylanilinium ion substrates in water are es-
sentially the same (ref 16), the second-order rate constants
for the reaction of methylpyridinium ions with iodide in
acetonitrile (ref 24) are significantly lower than the rates for
the NaN3 reaction in water (this study).
complex. While ∆Hq and ∆HR are lower for 1MeO,
presumably because of the resonance interaction, neither
1Me nor 1MeO shows any discernible pattern in the
change in charge at the reaction center (∆q) between the
initial and activated complexes for any Nu. For instance,
for 1Me ∆q ) 0.09, 0.22, 0.158, and 0.310 for I-, N3
,
-
Me3P, and water, respectively, values that do not cor-
relate with any computed energy.
The geometries of the activated complexes, however,
are anomalous. For the reactions of 1Me and 1MeO with
N3-, I-, and Me3P, the TS occurs at a C-Py bond length
of ∼1.75 Å (“early”) and an “early” or intermediate C-Nu
bond length. For the water reaction, however, the TS
occurs at C-O bond lengths of ∼1.8 Å for 1Me (“late”
for the Nu) and ∼2.25 Å for 1MeO (“early” for the Nu),
which is consistent with resonance stabilization of the
TS for 1MeO. The C-Py bond lengths are ∼2.15 Å
(“late” for the LG) in both cases. Thus, while the rates
are different, reflected in the different values of ∆Hq for
each reaction, the structures of the activated complexes
are the same for I-, N3-, and Me3P displacement, but not

7370 J . Org. Chem., Vol. 61, No. 21, 1996
Buckley and Oppenheimer
F igu r e 16. HOMOs in the activated complexes for the
reaction of water and I- with 1Me and 1MeO. The HOMO is
localized on the LG for the water reaction and on the Nu for
-
the iodide reaction; N3 and Me3P show essentially the same
pattern as I-. This difference appears to be related to the
position of the activated complex on the reaction surface.
F igu r e 17. Plots of the energy of the HOMO-LUMO gap vs
∆HR, a measure of the thermodynamic characteristics of the
reaction that shows the same anomaly as the cross-correlation
plots (Figure 11). Left to right the points are for I-, N3-, Me3P,
and water. The slopes are essentially the same, 0.086 (r )
0.999 97) for 1Me and 0.082 (r ) 0.9969) for 1MeO.
the fact that the activated complexes for the water
reactions are close in structure to the initial complexes
for the back reactions, pyridine displacement of water
from +H2OMe or +H2OCH2OMe. Thus, both charge
repulsion in the activated complex (for 1MeO) and the
position of the TS on the potential energy surface (for
1Me and 1MeO) contribute to the anomalous bond
lengths in the activated complexes.
from these data that the Brønsted value is independent
of the mechanism of substitution with the same LG class,
with comparable or larger values obtained for the SN2
reactions of substrates that do (benzyls) or do not
(methyls) contain potentially stable carbenium ions.
What is so startling about all the above results is the
lack of a clear predictive pattern between the Brønsted
coefficient and mechanism in series with similar LGs in
terms of charge and pKa. The computational results
suggest that there is no single source of the Brønsted
coefficients, the values of which appear to depend on the
interplay of several factors, and that they depend on the
system. Attempting to compare results between series
of different structures emphasizes Hammond’s specific
warning32a that free energies or their relations may not
correlate with the potential energy surfaces upon which
his postulate is based and that the postulate should not
be applied to systems in which the reactants and acti-
vated complexes are separated by large energies. As
Arnett and Reich27 point out, Hammond32a explicitly
excluded use of the postulate to evaluate SN2 reactions
of this type.
Th e P ossibility of a P r ea ssocia tion Mech a n ism .
In our study of the nucleophilic substitution reactions
of (4-methoxybenzyl)dimethylsulfonium chloride,10 we
found kinetic evidence for the simultaneous operation of
three substitution mechanisms: a stepwise mechanism
with capture of the carbenium ion by nucleophile and
solvent, a simple concerted displacement reactionsthe
Hughes-Ingold SN2 mechanismsand a preassociated-
concerted mechanism in which a negativesbut not a
neutralsnucleophile and positively charged substrate
With a different potential carbenium ion in the sub-
strate, the picture is also complicated and shows that âLG
values alone are not a good criterion with which to assess
mechanism. âLG values of -1.11 and -0.91 have been
reported for the hydrolysis (100 °C) and NAD+ gylcosyl
hydrolase-catalyzed cleavage (37 °C), respectively, of
NAD+ analogs with different pyridine LGs,4 and Sinnott33
reported a âLG value of -1.29 for the hydrolysis of â-D-
galactopyranosylpyridinium ions. While these values are
in the same high ranges as the clean concerted displace-
ment reactions for the methyl and benzyl substrates,
there are other results that suggest the ribosyl and
glycosyl reactions are dissociative. Hydrolysis reactions
for all the glycosyl-, and ribosyl-, and arabinosylpyri-
dinium ions have high, positive values of ∆Sq.33 Ta-Shma
and Oppenheimer34 found that that N3- traps the solvent-
separated NAD+ IDC, and in a recent study, Bennet35
found that products from the hydrolysis of deoxygluco-
sylpyridinium ions arise from trapping of the same
intermediate. These values are consistent with a “loose”
activated complex for a unimolecular process through a
stable oxocarbenium ion intermediate or IDC. It is clear
(33) Hosie, L.; Marshall, P. J .; Sinnott, M. L. J . Chem. Soc., Perkin
Trans. 2, 1984, 1121-1131. For earlier work, see: Capon, B. Chem.
Rev. 1969, 69, 407-498. Cordes, E. H.; Bull, H. G. Chem. Rev. 1973,
73, 581-603.
(34) Ta-Shma, R.; Oppenheimer, N. J ., Manuscript in preparation.
(35) Huang, X.; Surry, C.; Hiebert, T.; Bennet, A. J . J . Am. Chem.
Soc. 1995, 117, 10614-10621.

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 21, 1996 7371
F igu r e 19. Plots of kobsd vs [NaN3] with no control of ionic
strength by addition of exogenous salt. The circles are the
measured values, the ×’s values computed using a Winstein-
Fainberg b value of -0.4 determined from second-order rate
constants obtained under conditions of ionic strength. The line
through the ×’s has no physical meaning, but is there only to
guide the eye. The experimentally determined values appear
to match, within error, the computed line.
merely reflect the effect of ionic strength on rate and not
a preassociation mechanism. A negative primary salt
effect is evident even under conditions of constant ionic
strength, as shown by the slight curvature of the plot of
k
obsd vs [NaN3] at µ ) 1.7 (see Figure 1; similar curvature
is found in a plot of data for reaction at µ ) 4), which is
consistent with this view. In any case, because there is
no ROH formed for 1a , it is not possible to evaluate this
possibility with the definitive product ratio test used in
the sulfonium ion work.10
F igu r e 18. Plots of the energy of the HOMO for the activated
complex vs ∆Hq, a measure of the kinetic characteristics of
the reaction that shows the same anomaly as the cross-
correlation plots (Figure 11). Left to right the points are for
I-, N3-, Me3P, and water. The slopes are different because of
the differences in ∆Hq.
If there is no preassociation mechanism at work here,
it is not surprising. Pyridinium ions are well-known to
bind anions to the face of the pyridine ring to form
charge-transfer complexes, the basis of Kosower’s Z-scale
for solvent polarity.37 In these complexes, the nucleophile
would be bound nonproductively with regard to the
necessary geometry for the displacement reaction. In
addition, the charge on the ring is reduced by formation
of the charge-transfer complex, which makes the pyridine
a worse LG. None of these physical impediments is
preform an ion pair Nu:-‚RX+ that is an intermediate on
the reaction coordinate. In the presence of exogenous salt
added to control ionic strength, formation of the nucleo-
phile-substrate ion pair is in competition with formation
of the salt-substrate ion pair and is of much less
importance at lower than at higher [Nu]. In the presence
of added nucleophile only, however, the two concerted
mechanisms can compete efficiently. These effects could
be seen in plots of kobsd vs [Nu]; in the presence of
exogenous salt, this plot was linear over most of the range
of [Nu] (up to 5 M NaN3). With no added salt, however,
the plot exhibited a break at low [Nu] that signaled the
switch from Hughes-Ingold SN2 to preassociated-con-
certed mechanisms, with some diminution in the rate of
the unimolecular process related to the total ionic strength
of the solution. While the kinetic profile was clear, the
different mechanisms were established with certainty by
fitting the rate constants for each process to equations
that relate the product ratio [RNu]/[ROH] to the rate
constants for the various processes.10
present in Nu-‚RSMe2 ion pairs.
+
Th e Effect of Solven t. Our results for the pyri-
dinium and sulfonium ions10 are in accord with the
suggestion of Arnett and Reich24 that the effects of
solvation of the LG at the TS are important determinants
of reactivity. On the basis of the pKa’s of the LGs, it
would be expected that the dimethylsulfoniums (pKa )
-6) would react much faster than the pyridinium ions
(pKa ) 1.45-5.25). There is a difference in ∆Hq between
the series, with ∆Hq ∼ 22 kcal/mol for the sulfonium
ions10 and ∼16 kcal/mol for the pyridinium ions (Table
2), which reflects the difference in pKa and/or the stability
of the substrate as a Lewis complex (sulfonium ions:
soft-soft; pyridinium ions: soft-intermediate for the
carbenium ion-LG bond). Despite these differences, the
second-order rate constants at 80 °C are the same within
In principle, there is no reason why these pyridinium
ion substrates should not have the same mix of mecha-
nisms. As shown in Figure 19, the rate constants
obtained for the reaction of 1a with NaN3 alone in the
range 0-1.7 M NaN3 describe a smooth curve with rates
greater than found under constant ionic strength. If
there is a break in this curve, it is much too small to
detect accurately, unlike the results for the sulfonium
substrate. In addition, the points seem to fit fairly well
(36) Fainberg, A. H.; Winstein, S. J . Am. Chem. Soc. 1956, 78, 2763-
2767. In this instance we used the equation k ) ko (1 - b∆[salt]),
where ∆[salt] is the difference for reaction at two different ionic
strengths. Values of kcalc obtained are second-order rate constants at
the ionic strength of the nucleophile alone; computed kobsd are then
kcalc[Nu].
rate constants calculated with a Winstein-Fainberg36
b
value of -0.41 (x’s and line, Figure 19), which suggests
(37) Kosower, E. M. An Introduction to Physical Organic Chemistry;
Wiley: New York, 1968; pp 295-304.
that the rate constants in the absence of added salt

7372 J . Org. Chem., Vol. 61, No. 21, 1996
Buckley and Oppenheimer
a factor of 2. Thus, other factors must be important,
including the effect of solvent.
analogs: at the same temperature and pH, the enzyme-
catalyzed reaction is ca. 104 faster (in Vmax) than the
solution reaction.
The values of ∆Gq (25.2 kcal/mol) for the reaction of
1a and the corresponding dimethylsulfonium with NaN3
are the same, but the value of ∆∆Sq is -15 gibbs/mol.10
The enthalpy and entropy required to desolvate the
nucleophile and the translational entropy loss in bringing
the nucleophile and substrate together will be the same
for both substrates. For charged substrates, the ground
state and activated complexes both bear a full formal
positive charge, and only small adjustments in solvation
at the reaction center are needed as the transition state
is achieved with a subsequent small loss of solvation
entropy. In fact, solvolysis of (4-methoxybenzyl)dimeth-
ylsulfonium ion is not very sensitive to solvent polarity
and exhibits a low m value.11 The substantial difference
in ∆Sq must reflect the large, unfavorable change in
solvation entropy of introducing the hydrophilic 3-cyan-
opyridine into bulk solvent at the TS and the minor
entropic effects of introducing the hydrophobic SMe2 into
the cavities or interstitial spaces in bulk solvent. Sub-
stantial differences between ∆Sq for the SN1 solvolysis
How does an enzyme “catalyze” an SN1 reaction?
Binding of NAD+ in the enzyme active site should effec-
tively desolvate the pyridinium ion, and all other things
being equal, the difference in rate for bond cleavage may
be related to removal in the enzyme active site of the
retarding effects of an unfavorable ∆Sq caused by solva-
tion of the LG. The ∆∆Sq of -15 gibbs/mol between 1a
and the corresponding dimethylsulfonium ionsbetween
extensive solvent rearrangement for the pyridinium ion
and essentially no solvent reorganization for the di-
methylsulfonium ionsis worth ca. 3 × 103 in relative rate,
within a factor of 10 of the difference between the solution
and enzyme-catalyzed rates for NAD+ analogs. Thus, in
this system enzyme “catalysis” of bond cleavage may be
a solvent effect. This may not be the entire story,
however, because several 2′-substituted arabio NAD+
analogs are quite good inhibitors of the enzyme,40 which
is not consistent with a simple desolvation argument.
Reasons for the difference between the ribo and arabino
series are under active investigation.
of (4-methoxybenzyl)dimethylsulfonium in H2O (∆Sq
)
14.5 gibbs/mol)11 and in D2O (∆Sq ) 7 gibbs/mol)10 are
also consistent with this expression of the hydrophobic
effect on kinetics. Values of ∆S° are slightly more
negative for the introduction of noble gases and simple
hydrocarbons into D2O than into H2O,38 which is the
result of the slightly (5-7%) stronger deuterium bonds
in D2O. Thus, it is possible that 1a does not undergo an
SN1 reaction because of the unfavorable entropy of
solvation of the LG, but this is an unproven speculation
based on a rather large extrapolation.
The presence of a stable oxocarbenium ion intermedi-
ate in the enzyme active site, proposed earlier by Schu-
ber4 for glycosylhydrolases and confirmed by the recent
results of Richard39 for â-galactosidase, challenges the
contention18 that ribosyl- and glycosyloxocarbenium ions
cannot exist as intermediates in enzymes such as
lysozyme. We argued elsewhere8 that it is not always
wise to assign mechanism based on an extrapolation from
the gas to solution phase; these results suggest that
extrapolations from solution to enzymes must be done
with caution.
Differ en ce betw een th e Solu tion a n d En zym e-
Ca ta lyzed Clea va ge of th e Nicotin a m id e-Ribosyl
Bon d in NAD+. Handlon et al.3 have shown recently
that the Taft plots for hydrolysis (log kobsd) and enzyme-
catalyzed (log Vmax) cleavage of a series of 2′-substituted
â-nicotinamide ribosides and the corresponding 2′-
substituted NAD+ analogs give high FI values (-6.9 and
-9.5, respectively). The higher value for the enzyme
reaction may reflect different intermediates formed in
dissociative reactions, a solvent-separated IDC in solution
and an oxocarbenium ion bound to the active site of the
enzyme. On the basis of high âLG values and the
difference in partitioning of MeOH and water in the
products of solvolysis (kMeOH/kHOH ∼ 2) and the enzyme-
catalyzed reaction (kMeOH/kHOH ∼ 100), Schuber4 argued
that the enzyme-bound intermediate was a stable ribo-
syloxocarbenium ion. We reported recently8,14 that in the
gas phase the arabinosyloxocarbenium ion is more stable
than the (4-methoxybenzyl)carbenium ion. All of these
results are consistent with a recent conclusion of Richard
et al.39 that the intermediate formed by â-galactosidase-
catalyzed hydrolysis of galactosides is a stable oxocar-
benium ion in the active site. The results are not
consistent with a displacement mechanism involving an
active site nucleophile with formation a ribosyl-acyl
intermediate. There is a large rate difference, however,
between the solution and enzyme reactions for NAD+
Ack n ow led gm en t. Supported in part by NIH Grant
No. GM-22982 (N.J .O), a Biotechnology Grant from the
State of California (N.J .O., N.B.), and a Research Award
from the UCSF Graduate Division (N.B.). The UCSF
Magnetic Resonance Laboratory is supported in part by
grants from the NSF (DMB 8406826) and the NIH (RR-
01668 and RR-4789). Funds for the purchase and
support of the tandem spectrometer in the UCSF Mass
Spectrometry Facility (A. L. Burlingame, Director) were
provided by a grant from the Division of Research
Resources (RR 01614) of the NIH and a grant from the
NSF (DIR 8700766); we thank David Maltby for obtain-
ing the LSIMS spectra.
Su p p or tin g In for m a tion Ava ila ble: Full potential en-
ergy surfaces for the water and iodide displacement reactions
on methyl- and (methoxymethyl)pyridinium ions and various
correlation plots that show the correspondence between pyri-
dine and dimethylaniline leaving groups (7 pages). This
material is contained in libraries on microfiche, immediately
follows this article in the microfilm version of the journal, and
can be ordered from the ACS; see any current masthead page
for ordering information.
J O960729J
(39) Richard, J . P.; Westerfeld, J . G.; Lin, S.; Beard, J . Biochemistry
1995, 34, 11713-11724.
(40) Schuber, F.; Oppenheimer, N. J . Unpublished results.
(38) Ben-Naim, A. Solvation Thermodynamics; Plenum Press: New
York, 1987; pp 75-77.