Chemical models of peptide formation in translation

Article abstract of DOI:10.1021/bi1000273

Ribosomal incorporations of N-alkyl amino acids including proline are slower than incorporations of non-N-alkyl L-amino acids. The chemical reactivity hypothesis proposes that these results and the exclusion of nonproline N-alkyl amino acids from the gene

Full text of DOI:10.1021/bi1000273

Biochemistry 2010, 49, 2177–2185 2177
DOI: 10.1021/bi1000273
Chemical Models of Peptide Formation in Translation†
R. Edward Watts and Anthony C. Forster*
Department of Pharmacology and Vanderbilt Institute of Chemical Biology, Vanderbilt University Medical Center,
2222 Pierce Avenue, Nashville, Tennessee 37232
Received September 22, 2009; Revised Manuscript Received February 5, 2010
ABSTRACT: Ribosomal incorporations of N-alkyl amino acids including proline are slower than incorporations
of non-N-alkyl -amino acids. The chemical reactivity hypothesis proposes that these results and the exclusion of
L
nonproline N-alkyl amino acids from the genetic code are explained by intrinsic chemical reactivities of the
amino acid nucleophiles. However, there is little data on the reactivities relevant to physiological conditions.
Here, we use nonenzymatic, aqueous-based, buffered reactions with formylmethionine-N-hydroxysuccinimide
ester to model 11 amino acid nucleophiles in dipeptide formation. The relative rates in the nonenzymatic and
translation systems correlate well, supporting the chemical reactivity hypothesis and arguing that peptide bond
formation, not accommodation, is rate limiting for natural Pro-tRNAPro isoacceptors. The effects of N-sub-
stitution sterics, side chain sterics, induction, and pKa were evaluated in the chemical model. The dominant factor
affecting relative rates was found to be N-substitution sterics.
Ordinary natural substrates for translation encompass 19
amino acids and one N-alkyl amino acid, proline, esterified to
a larger number of tRNAs that recognize 61 sense codons.
Ribosomes thus utilize some 612 different codon pairs to template
the formation of some 400 different dipeptide bonds (and many
more dipeptides when supplied with unnatural amino acids).
Despite this enormous diversity of natural chemical couplings, it
is presumed that any codon pair can direct efficient coupling
between cognate amino acids if the substrate concentrations are
not limiting. The uniform decoding rate hypothesis (1) goes
further by predicting that all natural couplings occur at the same
rate at saturation. This hypothesis has been verified by measuring
in vitro rates of dipeptide formation from fMet-tRNAifMet and
nine amino acids (Phe, Ala, Gly, Ile, Val, Glu, Leu, Lys, and
His) delivered by 10 of the 46 aminoacyl-tRNAs of E. coli at
10 different codons at the ribosomal A site (1). The generally
believed mechanism of uniformity (1) is evolutionary tuning and
the masking of peptide bond formation rates by a much slower
accommodation step (movement of the A-site substrate into the
peptidyl-transferase center) (2). This rate-limiting accommoda-
tion hypothesis is based on the observation that accommodation
is rate-limiting in dipeptide formation when the A site substrate is
Phe-tRNAPhe chemically labeled with a large fluorescent mole-
cule (2). However, it should be noted that experimental chal-
lenges have prevented testing the rate-limiting accommodation
hypothesis with other substrates.
incorporation efficiencies of unnatural N-alkyl amino acids
(3, 4); rates of dipeptide formation with natural fMet-tRNAifMet
using a very efficient synthetic tRNAPhe-based adaptor at the
ribosomal A site were Phe>Ala>Pro>N-Me-Phe>N-Bu-Phe
((3); structures in Figure 1; summary in Figure 6 below). This
order of rates was correlated with the theoretical intrinsic
chemical reactivities of the amino nucleophiles, a chemical
reactivity hypothesis (4). Consistent with this hypothesis, the
rate-limiting step for all but our fastest coupling amino acid (Phe)
occurred after GTP hydrolysis by the aminoacyl-tRNA carrier,
elongation factor Tu (EF-Tu) (3). This narrowed down the steps
responsible for the large difference in rates to (i) substrate re-
lease from EF-Tu.GDP, (ii) substrate accommodation, and/or
(iii) peptide bond formation. It is unclear as to how our amino
acid substitutions could cause such differences in accommoda-
tion, as might be predicted by the rate-limiting accommodation
hypothesis. Alternatively, corresponding differences in peptide
bond formation rates would be predicted by the chemical
reactivity hypothesis. It therefore seems that there are at least
some natural exceptions to the uniform decoding rate hypothesis
and the rate-limiting accommodation hypothesis (the E. coli
Pro-tRNAPro isoacceptors), and that there are many unnatural
exceptions (N-alkyl aminoacyl-tRNAs).
The chemical reactivity hypothesis could also explain a long-
standing mystery regarding the origin of the genetic code (4) in
the RNA world. N-alkyl amino acids are major products of
prebiotic synthetic experiments and meteorite analyses; therefore,
why did Nature select Pro as the only N-alkyl amino acid in the
genetic code (5)? The chemical reactivity hypothesis proposes
that Pro was selected because it is the most reactive nucleophile
of the N-alkyl amino acids under physiological conditions.
Although the chemical reactivity hypothesis makes specific
predictions about the relative reactivities of the amino acid
nucleophiles on the ribosome and in the RNA world, there is
little data on the reactivities relevant to physiological conditions.
Prior comparisons of the nucleophilicities of various amino acids
and N-alkyl amino acids in nonenzymatic peptide bond for-
mation (aminolysis) were done mainly to optimize amino acid
Recently, we found that natural E. coli Pro-tRNAPro isoac-
ceptors provide interesting exceptions to the uniform decoding
hypothesis (3). These substrates couple 3- to 6-fold slower than
natural Phe-tRNAPhe to fMet-tRNAifMet in vitro, depending on
the Pro codon (<15% standard errors). The uniform decoding
hypothesis also does not explain our dramatically different
†This research was supported by the National Institutes of Health and
the American Cancer Society.
*To whom correspondence should be addressed. Department of
Pharmacology, Vanderbilt University Medical Center, 2222 Pierce
Ave., Nashville, TN 37232. Phone: (615) 936 3112. Fax: (615) 936 5555.
E-mail: a.forster@vanderbilt.edu.
r
2010 American Chemical Society
Published on Web 02/08/2010
pubs.acs.org/Biochemistry

2178 Biochemistry, Vol. 49, No. 10, 2010
Watts and Forster
MATERIALS AND METHODS
Materials. Formyl-methionine and N-Me-glycinamide hydro-
chloride were purchased from Bachem Americas, Inc. (Torrance,
CA) and 3 mm NMR tubes from Bruker Biospin (Billerica, MA).
All other compounds, unless noted below, were purchased from
Sigma-Aldrich, Inc. (St. Louis, MO).
Synthesis of Reagents. Formyl-methionine-hydroxy-
succinimidyl Ester (fMet-NHS). Formyl-methionine (1240.5
mg, 7 mmol) was dissolved in 75 mL of CH2Cl2. Dimethylami-
nopyridine (43 mg, 0.35 mmol, 0.05 equiv), N-(3-dimethyl-
aminopropyl)-N0-ethylcarbodiimide hydrochloride (2.7 g, 14 mmol,
2 equiv), and N-hydroxysuccinimide (1.6 g, 14 mmol, 2 equiv)
were added sequentially and stirred overnight at room tempera-
ture. The reaction mixture was concentrated under reduced
pressure to ∼25 mL and loaded onto a silica flash column
solvated with CH2Cl2. It was eluted with 0-30% acetone in
CH2Cl2, concentrated under reduced pressure, dried with toluene
azeotrope, and the resultant oil dried under high vacuum for 48 h.
Yield: 1.721 g, 6.27 mmol, 89.6%. 1H NMR (400 MHz, CDCl3):
δ 2.122 (3H, s), 2.177 (2H, m), 2.647 (2H, m), 2.864 (4H, bs),
5.222 (1H, dddd, J=5.2, 7.6 Hz), 6.604 (1H, d, J = 8.0 Hz), 8.267
(1H, s). 13C NMR (100 MHz, CDCl3): δ 15.32, 25.33, 29.41,
31.51, 48.38, 160.75, 167.30, 168.54. High resolution mass
spectrometry (HRMS) m/z calculated for C10H14N2O5S,
274.0623; found, 297.0518 (m þ Naþ). The pKa of the NHS
leaving group is 5.9 (15).
N-Methyl-phenylalaninamide. N-Methyl phenylalanine
(400 mg, 2.23 mmol) was dissolved in 17.1 mL of 2:1 dioxane/
water. 9-Fluorenylmethyl chloroformate (577 mg, 2.23 mmol,
1 equiv) and K2CO3 (616.4 mg, 4.46 mmol, 2 equiv) were added.
After 24 h, TLC in 7% acetone and 0.5% HOAc in CH2Cl2
showed very little 9-fluorenylmethyl chloroformate present
(Rf = 0.6) and a preponderance of product (Rf = 0.5) when
visualized by UV. Staining the plate with ninhydrin showed no
amine present in the reaction mixture, whereas a known sample
of N-methylphenylalanine stained red on the baseline. Staining
with bromocresol green showed the presence of a carboxylic acid
with an Rf = 0.5, corresponding to the Rf of the product, but no
carboxylic acid on the baseline. The reaction was acidified and
brine (15 mL) added before extraction (4 ꢀ100 mL) with EtOAc
and solvent removal with reduced pressure. This reaction was
repeated exactly to increase the amount of total product avail-
able. Without further purification, 1.0 g of the Fmoc-N-methyl-
phenylalanine was dissolved in 25 mL of dry THF, and triethy-
lamine (382 μL, 2.74 mmol, 1.1 equiv) was added. This mixture
was cooled in an ice/water bath, whereupon ethylchloroformate
(261 μL, 2.74 mmol, 1.1 equiv) was added in aliquots over 2 min.
This was stirred in the ice/water bath for 1 h, then ammonium
hydroxide (1.18 mL, an excess) at room temperature was added
in one aliquot to the cold reaction mixture with vigorous stirring.
The reaction was allowed to warm to room temperature and was
stirred overnight. The reaction was diluted with water and
extracted with EtOAc (4 ꢀ 50 mL). These combined organic
phases contained Fmoc-protected N-methyl phenylalaninamide.
The combined aqueous phases were basified with K2CO3 and
extracted with EtOAc (4 ꢀ 50 mL). These combined organic
phases contained N-methyl phenylalaninamide that had been
deprotected by the basic conditions of the reaction. The com-
bined organic phases from both extractions were concentrated
under reduced pressure, and 10 mL of 20% piperidine in DMF
was added and reacted for 24 h. The DMF/piperidine was
F
in physiological-like buffer.
IGURE 1: Amino acid amide nucleophiles reacted with fMet-NHS
couplings in solid phase peptide synthesis (e.g., Gly=Pro>Ala>
Phe in ref 6; Ala> Gly>Phe>Pro in ref 7). These reactivity
orders differ from our expected relative chemical reactivities
of the amino nucleophiles under physiological conditions (Phe>
Ala>Pro (3)). However, such studies are of limited relevance to
translation because they used activated intermediates and com-
plex multistep mechanisms (e.g., carbodiimide and hexafluoro-
phosphate 2-(7-Aza-1H-benzotriazole-1-yl)-1,1,3,3-tetramethyl-
uronium (HATU1) chemistries) that are very different from
aminoacyl-tRNA ester chemistry, and reactions have been
performed almost exclusively in anhydrous organic solvents to
prevent hydrolysis. Anhydrous solvents are problematic for
modeling translation because pH is not well-controlled, and
pKa values can differ from aqueous solution by many units (8).
pH is an important variable because the % protonation of the
amine nucleophile varies considerably between different amino
acids under physiological conditions ((9) Table S3 (Supporting
Information)), and deprotonation is necessary for nucleophilic
attack. Though it is possible, if not likely, that the amine is not
fully accessible to the bulk solvent on the ribosome (10) and its
pKa is shifted, even deprotonation of the ammonium form by a
proton shuttle involving RNA (11) should be affected by amine
pKa. Indeed, the rates of peptide formation by puromycin
analogues at the ribosomal peptidyl transferase center under
physiological conditions exhibit dependence on pH, and this pH
dependence is altered by changing the pKa of the amine nucleo-
phile (12). These data and considerations argue that aqueous
buffers at near-neutral pH are preferable for modeling peptide
bond formation (13) in translation. Yet, the only relevant such
rate comparison that we are aware of showed Gly>Ala>Pro>
Phe (14). This order also differs from our expected relative
chemical reactivities, although the relevance of this study to
translation was limited by using activated acetates rather than
activated amino acids as electrophiles, by using phosphate as the
ester leaving group (which is much less hindered sterically than
the ribose of tRNA), and by not testing N-alkyl amino acids
other than Pro. Thus, we examine here alternative leaving groups
under physiological-like conditions in order to model the nucleo-
philicities of amino acids studied on the ribosome.
1Abbreviations: AA, amino acid; ACN, acetonitrile; HATU, hexafluoro-
phosphate 2-(7-Aza-1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium;
HRMS, high resolution mass spectrometry; NHS, N-hydroxysuccini-
mide; UPLC-MS, ultrahigh pressure liquid chromatography-mass
spectrometry.

Article
Biochemistry, Vol. 49, No. 10, 2010 2179
removed under reduced pressure, and the material purified by
m/z calculated for C9H17N3O3S, 247.0991; found, 270.0883
(m þ Naþ).
flash chromatography in 0-10% methanol in 3% triethylamine/
CH2Cl2. Yield: 286 mg, 1.6 mmol, 64%. H NMR (400 MHz,
Formyl-methionine-phenylalaninamide. 1H NMR (600 MHz,
30% CD3CN/70% D2O): δ 1.908 (2H, ddd, J = 7.2 Hz), 2.224
(3H, s), 2.304 (2H, m), 3.093 (1H, dd, J = 11.4, 14.4 Hz), 3.549
(1H, dd, J = 4.8, 14.4 Hz), 4.579 (1H, dd, J = 7.2 Hz), 4.887 (1H,
dd, J = 4.8, 11.4 Hz), 7.530 (5H, m), 8.282 (1H, s). 13C NMR
(150 MHz, 30% CD3CN/70% D2O): δ 11.46, 25.99, 28.01, 34.24,
48.92, 51.79124.55, 126.16, 126.48, 134.31, 161.19, 169.98,
172.89. HRMS m/z calculated for C15H21N3O3S, 323.1304;
found, 346.1196 (m þ Naþ).
1
CD3CN): δ 2.305 (3H, s), 2.751 (1H, dd, J = 8.4, 13.6 Hz), 2.85
(2H, bs), 3.027 (1H, dd, J = 5.2, 13.6 Hz), 3.162 (1H, dd, J = 5.2,
8.4 Hz), 6.344 (1H, bs), 7.063 (1H, bs), 7.223 (5H, m). 13C NMR
(100 MHz, CD3CN): δ 35.98, 40.70, 67.53, 127.83, 129.80,
130.74, 140.22, 176.86. HRMS m/z calculated for C10H14N2O,
178.1106; found, 179.1181 (m þ Hþ).
N-Butyl-phenylalaninamide. Phenylalaninamide (492.6 mg,
3 mmol) and butanal (537 μL, 6 mmol) were added to 15 mL of
methanol at room temperature and stirred 15 min. The reaction
vessel was cooled in an ice/water bath for 15 min, whereupon
NaBH4 (158.9 mg, 4.2 mmol) was added in 4 equal aliquots at
5 min intervals. The reaction was allowed to stir and warm to
room temperature overnight. TLC in 5% acetone and 5% triethyl-
amine in dichloromethane showed starting material and a product
with the ninhydrin stain. The reaction mixture was cooled again in
an ice/water bath, 537 μL more butanal was added, and stirred for
15 min, whereupon 158.9 mg more NaBH4 was added in 4 aliquots
at 5 min intervals. The reaction was allowed to stir and warm to
room temperature over 24 h. TLC showed greatly diminished
starting material relative to the product. The crude was purified by
flash chromatography with a 0-10% gradient of methanol in a
mixture of 10% ethyl acetate and 2% triethylamine in dichloro-
methane. Solvents were removed by reduced pressure. Yield:
Formyl-methionine-prolinamide. 1H NMR (600 MHz,
30% CD3CN/70% D2O): δ 2.163 (2H, m), 2.232 (1H, m),
2.306 (2H, m), 2.371 (3H, s), 2.806 (2H, m), 3.930 (1H. m),
4.059 (1H, m), 4.611 (1H, dd, J = 6.0, 8.4 Hz), 5.119 (1H, dd, J =
4.8, 9.0 Hz), 8.338 (1H, s). 13C NMR (150 MHz, 30% CD3CN/
70% D2O): δ 11.76, 22.06, 26.64, 26.93, 27.61, 45.29, 46.30, 57.70,
160.95, 168.59, 173.78. HRMS m/z calculated for C11H19N3O3S,
273.1147; found, 296.1043 (m þ Naþ).
Formyl-methionine-N-Me-glycinamide. 1H NMR (600
MHz, 30% CD3CN/70% D2O): δ 2.134 (2H, m), 2.351 (1H,
s), 2.742 (2H, m), 3.448 (3H, s), 4.24 (2H, dd, J = 16.8, 43.2 Hz),
5.298 (1H, dd, J = 4.2, 9.0 Hz), 8.329 (1H, s). 13C NMR (150
MHz, 30% CD3CN/70% D2O): δ 11.72, 26.64, 27.68, 34.14,
44.80, 48.54, 160.81, 169.65, 170.24. HRMS m/z calculated for
C9H17N3O3S, 247.0991; found, 270.0884 (m þ Naþ).
1
Dipeptide Formation Kinetics. An electrophile-containing
solution was added to an excess of nucleophile-containing
solution with vigorous stirring. Aliquots of the reaction mixture
were removed and added to a quench buffer at recorded intervals.
287.3 mg, 1.3 mmol, 43.5%. H NMR (400 MHz, D6-acetone):
δ 0.804 (3H, dd, J = 4.8 Hz), 1.198 (2H, m), 1.306 (2H, m), 2.390
(1H, ddd, J=4.4, 8.8 Hz), 2.501 (1H, ddd, J = 4.8, 9.6 Hz), 2.715
(1H, dd, J=5.6, 9.2 Hz), 3.009 (1H, dd, J = 3.2, 9.2 Hz), 3.206
(1H, dd, J = 3.6, 6.0 Hz), 6.25 (1H, bs), 7.05 (1H, bs), 7.189
(5H, m). 13C NMR (100 MHz, D6-acetone): δ 13.20, 19.89,
31.95, 39.20, 47.90, 64.01, 126.21, 128.17, 129.11, 138.62, 175.67.
HRMS m/z calculated for C13H20N2O, 220.3107; found, 221.1650
(m þ Hþ).
1
The quenched samples were diluted and analyzed by H NMR
(e.g., Figure 2A). The electrophilic solution contained fMet-NHS
and pyrazine (as an internal NMR integration standard) in 80%
deuterated DMSO and 20% deuterated acetonitrile (ACN). The
DMSO was necessary to effect rapid dissolution of the fMet-
NHS in the aqueous nucleophilic buffer. The ACN was added to
the DMSO solution immediately prior to use to keep the
electrophilic solution from freezing at the reaction temperature
of 4 ꢀC. The final fraction of organic solvent in D2O was 7.5%.
The nucleophile-containing buffer comprised varying concentra-
tions of amino acid amide hydrochloride salts in D2O and
potassium phosphate buffer. The nucleophilic buffer was ad-
justed to a reading of 7.0 on the pH meter by the addition of
KOD. The pD of a solution with a reading of 7.0 on the pH meter
was calculated to be 7.4 (16). The ionic strength of the reactions
was kept constant across all concentrations of all nucleophiles by
adjusting the final [Cl-] with added NaCl. The starting concen-
trations of species in the reaction mixture were as follows: fMet-
NHS (6.7 mM); pyrazine (1.25 mM); KHPO4 buffer (40 mM);
NaCl (785 mM); the concentrations of amino acid amides varied.
High ionic strength was used because it substantially improved
assay reproducibility by greatly reducing hydrolysis upon
quenching. Changes in rates of aminolysis due to increased ionic
strength did not change the relative order of rates of the
nucleophiles. For each reaction, 13-16 time points of 100 μL
were collected and quenched by adding them to 35 μL of quench
buffer containing 300 mM hydroxylamine and 100 mM KHPO4
buffer in deuterium oxide at pD 7.4 (Figure 2A).
Synthesis and Characterization of Dipeptide Products.
One milliliter-scale reactions using conditions identical to those
of the kinetic assays (with the highest concentration of nucleo-
phile) were performed and allowed to run for >10 h without the
removal of aliquots or quenching. These samples were purified by
preparative HPLC, dissolved in 30% CD3CN/D2O, and ana-
lyzed by 1H NMR, 13C NMR, and high resolution mass spectro-
scopy. Aliquots of the purified, characterized compounds were
added to quenched time point samples saved from the kinetic
assays. In all cases, the putative product peaks in the time point
samples increased relative to the hydrolysis byproduct and
hydroxylamine-quenched starting material upon the addition
of the known standard compound. NMR data for the primary
rotamers are given.
Formyl-methionine-glycinamide. 1H NMR (600 MHz,
30% CD3CN/70% D2O): δ 2.175 (2H, m), 2.351 (3H, s), 2.694
(2H, m), 3.988 (2H, dd, J = 16.8 Hz), 4.677 (∼2H, m) (this signal
obscured by the water signal), 8.290 (1H, s). 13C NMR (150
MHz, 30% CD3CN/70% D2O): δ 11.49, 26.52, 27.65, 39.42,
48.81, 159.69, 161.66, 170.84. HRMS m/z calculated for
C8H15N3O3S, 233.0834; found, 256.0725 (m þ Naþ).
Formyl-methionine-alaninamide. 1H NMR (600 MHz,
30% CD3CN/70% D2O): δ 1.619 (3H, d, J = 7.2 Hz), 2.209
(2H, m), 2.356 (3H, s), 2.791 (2H, m), 4.522 (1H, ddd, J=7.2 Hz),
4.754 (∼1H, m) (this signal was obscured by the water signal),
8.377 (1H, s). 13C NMR (150 MHz, 30% CD3CN/70% D2O):
δ 11.59, 14.14, 46.74, 48.61, 161.46, 169.88, 174.51. HRMS
1
Samples were analyzed with 600 MHz H NMR (Bruker) at
298 K with T1 = 10 s. At least 16 transients were collected for
each time point. Sufficient resolution of the formyl proton
resonances of the different species present required the use of

2180 Biochemistry, Vol. 49, No. 10, 2010
Watts and Forster
420 μL with D2O and ACN. Achieving sufficient resolution in the
alaninamide and phenylalaninamide reaction samples required
adding ACN to quenched samples to give a final concentration of
30% deuterated ACN in D2O, whereas, the prolinamide, glycin-
amide, and N-Me-glycinamide reactions required 20% ACN.
Results were quantified by integrating resonances of the formyl
protons on the dipeptide product, free acid (hydrolysis) byproduct/
impurity, and hydroxylamine (quenched) adduct of the starting
material with reference to the integration of the pyrazine internal
standard. The data collected were fitted using equations for
competing reactions (e.g., Figure 2B).
Equations. For A þ B þ H2O f P þ D, where P is the
aminolysis product of starting material ester A and amine B, and
D is the hydrolysis product of A and water,
d½Aꢁ=dt ¼ -k1½Aꢁ½Bꢁ -k2½Aꢁ ðdisappearance of starting material AÞ
d½Pꢁ=dt ¼ k1½Aꢁ½Bꢁ ðappearance of aminolysis productÞ
d½Dꢁ=dt ¼ k2½Aꢁ ðappearance of hydrolysis productÞ
Since the rate of disappearance of the starting ester is equal to the
sum of the rates of appearance of the products of the aminolysis
and hydrolysis reactions, then:
kTotal ¼ k1 þ k2
Integrating the aminolysis product rate equation gives:
½Pꢁ ¼ ð -k1=ðk1 þ k2ÞÞ ꢀ ½Aꢁ ꢀ ½Bꢁ ꢀ ð1 -e-ðk1þk2Þt
Þ
0
0
where [A]0 is the initial concentration of the starting material
ester, and [B]0 is the initial concentration of the starting material
amine. Removing [B]0 from this equation gives:
½Pꢁ ¼ ð -kapparent=ðk1 þ k2ÞÞ ꢀ ½Aꢁ ꢀ ð1 -e-ðk1þk2Þt
Þ
ð1Þ
0
where kapparent is the rate constant for the aminolysis reaction that
is dependent on the initial amine concentration [B]0, and which
gives k1 as the slope of the line from the relationship:
F
IGURE 2: Representative time course of dipeptide synthesis. (A)
NMR chemical shifts of products from a reaction between 762 mM
total prolinamide and 6.7 mM fMet-NHS over time. Signals due to
the formyl proton of fMet-Pro (all rotamers) are at 7.99, 7.98, 7.97,
and 7.95 ppm. Residual fMet-NHS starting material was quenched
with hydroxylamine, yielding fMet-NHOH (formyl protons at and
above 8.00 ppm) and fMet-OH (resonances at 7.96 and 7.93).
Concentrations of species were determined by integrating resonances
with respect to the resonance of pyrazine, which was added to the
electrophilic solution as an internal standard (resonance of pyrazine at
8.55 ppm is not shown). Sixteen time points were collected at each
concentration, but only every other time point is shown for clarity. Note
the minimal formation of fMet-OH hydrolysis product beyond that
present in the starting material. Also, detection of the minor rotameric
signal on the right was variable because of high sensitivity to the exact
solvent concentrations. (B) Representative fit of eq 1 to the time course in
A. Concentration error bars were determined by measuring differences in
the ratio of the internal standard resonance area to the sum of all analyte
resonance areas (this ratio should be invariant between time points).
kapparent ¼ k1½Bꢁ0
ð2Þ
Fitting eq 1 to the data collected in kinetics assays performed at
varying concentrations of amine gave a set of values for kapparent that
were plotted against the concentration of amine using eqs 2 and 3.
2
kapparent ¼ k3½Bꢁ0 þ k4½Bꢁ0
ð3Þ
Equation 2 above assumes that a negligible amount of base catalysis
occurs under the conditions of the assays. Equation 3 governs the
reaction where the amino acid amide acts as a general base catalyst
(right of Tables S1 and S2 (Supporting Information)) and where the
reaction proceeds by both catalyzed and uncatalyzed mechanisms.
In eq 3, k3 represents the rate of the uncatalyzed reaction pathway,
and k4 represents the rate of the catalyzed reaction pathway. All
data were fit with regressions using GraphPad Prism 4. The
concentration of the unprotonated amine was calculated with the
Henderson-Hasselbalch equation.
3 mm NMR tubes and dilution with D2O to help overcome
tuning and shimming problems arising from the high concentra-
tions of salt. It was also necessary to add deuterated ACN to the
quenched samples to improve resolution. Resolution may have
been improved by lowering the viscosity of the solvent to allow
better molecular tumbling of analytes and/or by a chemical shift
effect from changes in pH upon the addition of the organic
solvent. All of the 135 μL quenched samples were diluted to
Relative Rates of Dipeptide Formations in Competition
Assays. To a D2O buffer containing 30 mM KHPO4 at pD 7.4,
100 mM NaCl, N-Me-glycinamide, and an amino acid nucleo-
phile was added a solution of fMet-NHS in deuterated DMSO to

Article
Biochemistry, Vol. 49, No. 10, 2010 2181
make a final fMet-NHS concentration of 30 mM and a final
concentration of DMSO of 10% v/v. The ratio of N-Me-
glycinamide to the amino acid nucleophiles varied depending
on the solubility of the amino acid nucleophile but typically
ranged from a 4-fold excess of N-Me-glycinamide to an equimo-
lar ratio for each nucleophile. The initial concentration of the
amino acid nucleophile was never below 3ꢀ that of fMet-NHS to
ensure that the reaction rates would remain relatively constant
with respect to both nucleophiles. Reactions were run for 24 h at
room temperature, diluted with D2O, and titrated with deuter-
ated ACN. NMR scans were taken between ACN additions until
adequate resolution of the fMet-AA dipeptide product and fMet-
OH hydrolysis product was achieved. Typically, the final con-
centration of ACN fell between 15 and 30%. NMR resonances
were assigned with the help of standards, and pyrazine in the
fMet-NHS sample served as an integration standard. Samples
were spun on a Bruker cryoprobe 600 MHz NMR at 410 K with
16 transients collected and a relaxation delay of 10 s. Relative
rates of nucleophile reactivity were taken to equal the relative
concentrations of dipeptide products at equilibrium adjusted by
the relative initial concentrations of nucleophiles in accordance
with the following equation:
fMet-Phe-NH2, C15H21N3O3S, 323.1304; found, 346.1201 (m þ
Naþ). HRMS m/z calculated for fMet-N-Me-Phe-NH2,
C16H23N3O3S, 337.1460; found, 360.1379 (m þ Naþ).
Determination of Amino Acid Amide pKa Values. Calcu-
lation of the concentration of unprotonated amine in the reaction
buffer required the measurement of the pKa values of the amino
acid amide nucleophiles under the solvent conditions of the
study: [NaCl] = 585 mM, [amino acid amide HCl] = 185 mM,
and 7% DMSO in H2O. Two hundred millimolar KOH was
added in measured aliquots and the pH measured with a Mettler
Toledo SevenEasy pH meter calibrated with standard buffer
solutions. For each titration, >40 observations were made.
Inflection points at the beginning and end of the titrations were
found by plotting the instantaneous slope of the pH observations
versus the amount of titrant added (Figure 3). The addition of
NaCl and DMSO had a minimal effect on pKa based on
comparisons with the literature (Table S3 (Supporting In-
formation) and ref 9). The inflection points appear as peaks,
and the pH at the midpoint between these peaks was taken to be
the pKa of the compound. Equation 4 (17) was used to calculate
pKa values in D2O (Table S3 (Supporting Information)).
pKa in D2O ¼ ðpKa in H2O -0:42Þ=0:929
ð4Þ
relative rate ¼ ð½fMet-AAeqꢁ
The relative pKa values of the amino acid amides correlate with
the relative pKa values of the amino groups of the amino acids
(pKa of Gly = 9.58, Phe=9.09, Ala=9.71, Pro=10.47, and
N-Me-Gly=9.97) reported in the CRC Handbook online.
Effect of pD on Reactivity. In a related chemical model for
dipeptide formation, the rate of aminolysis was shown to be
dependent on pD in a manner dictated by the concentration of
free amine, as expected (18). We used a simpler approach to
confirm the expected effect of unprotonated amine concentration
in our model by performing assays identical to our kinetics assays
except that the reactions were allowed to reach equilibrium. One
series of reactions varied concentrations of total amine at fixed
pD 7.4. The other series of reactions used a fixed concentration of
total amine and varied pD by the addition of small aliquots
(<0.5% of total reaction volume) of 3 M HCl in 90% D2O/water
or 3 M KOD in 99% D2O. In the fixed pD assays, the median
range in the concentration of unprotonated amine was 3.25-fold.
In the varying pD assays, the mean range in the concentration of
unprotonated amine was 3.5-fold, and the mean range in con-
centration of deuteroxide ion was ∼10-fold, from pD 7.19 to
pD 8.22. The reactions were allowed to run for 18 h (∼t = ¥) and
equilibrium concentrations of aminolysis and hydrolysis pro-
ducts quantified by NMR as per the kinetics assays. The fraction
of fMet-NHS converted to dipeptide via aminolysis correlated
with the concentration of unprotonated amine. The fractions of
ester converted by aminolysis were essentially the same regardless
of whether unprotonated amine concentrations were varied by
changing the concentration of total amine at constant pD or by
changing pD at constant concentration of total amine (Figure 4).
ꢀ ½N-Me-Gly0ꢁÞ=ð½fMet-N-Me-Glyeqꢁ ꢀ ½AA0ꢁÞ
where [fMet-AAeq] is the equilibrium concentration of the
dipeptide containing fMet and the test amino acid nucleophile,
[fMet-N-Me-Glyeq] is the equilibrium concentration of the di-
peptide containing fMet and N-Me-glycinamide, [N-Me-Gly0] is
the initial concentration of N-Me-glycinamide, and [AA0] is the
initial concentration of the test amino acid nucleophile. Relative
rates were constant across the concentration range for all amino
acid nucleophiles. As expected, the concentration of the fMet-
OH hydrolysis product was constant across all experiments with
a given amino acid nucleophile since the concentration of the
deuteroxide was held constant by the buffer.
Relative Rates of Dipeptide Formation from N-Me-Phe-
NH2 versus Phe-NH2. fMet-NHS hydrolyzes much faster than
it reacts with N-Me-Phe-NH2, preventing the detection of the
fMet-N-Me-Phe-NH2 product by NMR or UV absorbance.
However, in competition reactions between N-Me-Phe-NH2
and Phe-NH2 for coupling to fMet-NHS, both dipeptide pro-
ducts gave detectable mass spec extracted ion peaks, enabling the
estimation of the relative rates (assuming their similar structures
give similar relative ionization efficiencies). Five reactions con-
taining 2.5 mM fMet-NHS, 7.5 mM Phe-NH2, and amounts
of N-Me-Phe-NH2 that were equimolar or 5-fold, 10-fold,
50-fold, and 100-fold in excess over Phe-NH2 were conducted
in ∼400 mM H2O in DMSO. The reactions were run for 20 min at
90 ꢀC, then for 16 h at room temperature. Five hundred micro-
liters of 1 M phosphate buffer at pH 4.2 and 100 μL of brine were
added, followed by 4 mL of EtOAc. After vigorous stirring, the
organic layer was removed, dried with sodium sulfate, pushed
through a silica plug, and evaporated under reduced pressure.
Products were resuspended in 50 μL of ACN and then diluted to
150 μL with water. Samples were analyzed by ultrahigh pressure
liquid chromatography-mass spectrometry (UPLC-MS), the
identities of the dipeptide products in the reaction were confirmed
by HRMS, and the extracted ion peaks of the dipeptide product
masses were integrated. Plots of product ratios to starting
material ratios gave straight lines. HRMS m/z calculated for
RESULTS AND DISCUSSION
Our initial experiments were based upon a promising che-
mical model of dipeptide formation that used aqueous buffer at
physiological pH. The electrophile was formyl-Phe-trifluoro-
ethyl ester, and the nucleophile was glycinamide (Gly-NH2;
the amide neutralizes the unphysiological negative charge of
the carboxylic acid (18)). However, we could not adapt this
system readily to other amino acid amide nucleophiles because

2182 Biochemistry, Vol. 49, No. 10, 2010
Watts and Forster
F
IGURE 3: Measurements of pKa values of amino acid amide nucleophiles. (Left panel) pH titration curves. The pKa was taken to be the pH at
the midpoint of the titration. (Right panel) Plots of the instantaneous slope. The midpoint was determined by finding the inflection points of the
titration.
peptide bond formation was much slower than ester hydro-
lysis, even at the highest possible concentrations. Presumably
Gly-NH2 was the most reactive of the amino acid amides we
attempted because it is the least hindered sterically. Using the
trifluoroethyl ester of fMet instead of fPhe resulted in even
lower product yields. Schemes for continuous in situ activation
of fMet were also deemed impractical due to mechanistic
complexity.
Next, we turned to carboxylic acid activation by N-hydroxy-
succinimidyl (NHS) ester formation because of its utility in
labeling proteinogenic lysines in aqueous solution and because
its steric bulk is more comparable to that of ribose. fMet-NHS

Article
Biochemistry, Vol. 49, No. 10, 2010 2183
F
IGURE 4: Fractions of ester converted by aminolysis are unchanged regardless of whether the concentration of the unprotonated amine is varied
by changing the concentration of total amine or by changing pD. Triangles show the effect of varying the concentration of unprotonated amine at
a fixed pD. Circles show the effect of varying the concentration of unprotonated amine by changing the pD at a fixed concentration of total amine.
(Figure 1, top left) proved sufficiently soluble in aqueous buffer,
sufficiently activated as an electrophile, and sufficiently resistant
to hydrolysis to enable differentiation of its reaction rate with
four of the five nucleophiles we had used on the ribosome.
Because N-Me-Phe-NH2 and N-Bu-Phe-NH2 were unreactive,
the least sterically hindered amino and N-alkyl amino acids
possible, Gly-NH2 and N-Me-Gly-NH2, were added to our set
to model the effect of N-methylation (Figure 1).
The aqueous component of the buffer was 99.9% D2O to
enable quantification of reaction components by 1H NMR.
Though the kinetic isotope effect may reduce reaction rates,
it is expected to be too small (∼10%) to alter the order of
reactivity across the set of nucleophiles studied here. A small
amount of organic solvent (7.5%) was also necessary for
solubility and to prevent freezing; this had little effect on amine
pKa values (Table S3 (Supporting Information)), and even living
cells contain a significant concentration of organics, such as
lipids.
see right side of Table S1 (Supporting Information)). The order
of reactivity was Gly-NH2>Phe-NH2>Ala-NH2>Pro-NH2>
N-Me-Gly-NH2>undetectable N-Me-Phe-NH2 and N-Bu-Phe-
NH2. This order correlates well with the order of our ribosome-
catalyzed reactions when controlling for the effect of the tRNA
and codon by using the same tRNA/codon combination
(Figure 6; P<0.05%). Though N-Me-Phe-NH2 incorporation
was undetectable by our NMR assay, a mass spectrometry assay
enabled an approximation of its incorporation rate as 500-fold
slower than that of Phe-NH2 (Figure 6; see Materials and
Methods). This confirmed the effect of N-methylation observed
by comparing Gly-NH2 with N-Me-Gly-NH2 (630-fold;
Figure 5A and Table S1 (Supporting Information)). However,
the extents of rate changes with different amino acids in the
presence and absence of ribosomes did not correlate perfectly,
especially for N-methylation (Figure 6). Possible explanations
include nucleophile orientation preferences when bound to
tRNA and the peptidyl transferase center that are not experi-
enced in free solution and/or different chemical and steric effects
of the tRNA ester versus our NHS ester substituent. Our che-
mical model rates in Figure 5A also matched fairly well the ribo-
somal rates with different tRNAs and codons (cognate tRNAs
and their various codons: within experimental errors, those
The apparent rates of dipeptide formation were plotted against
the concentration of total amine (protonated and unprotonated)
to incorporate all effects without data manipulation, as is done
for ribosomal rates (Figure 5A). There was some curvature in the
plots due to base catalysis (detailed in Materials and Methods;

2184 Biochemistry, Vol. 49, No. 10, 2010
Watts and Forster
F
IGURE 6: Comparison between the relative rates of reaction in the
chemical model system with those using the tRNAPheB adaptor in the
ribosomal system. The ribosomal reaction rates (kacc,pep) do not
include the GTP hydrolysis rates of EF-Tu (kGTP) (3). Phe rates in
both systems are normalized to 100,000, and the other rates are
multiplied by the same factors and plotted on a log scale. Error bars
indicate the 95% confidence interval (2 standard deviations).
measured pKa values and the Henderson-Hasselbalch equation
(Figure 4). This indicated that pD changes affected aminolysis
much more than hydrolysis and supported the expectation that
aminolysis rates are dictated by the concentation of unproto-
nated amine. Replotting the apparent rates versus concentration
of unprotonated amine calculated from our pKa measurements
gives much larger rate constants in all cases (Figure 5B; Table S2
(Supporting Information), indicating that the chemical model
rates are dramatically affected by pH under physiological con-
ditions. However, the different pKa values of the amino and
N-alkyl amino acid amides do not correlate well with the relative
rates (Table S3 (Supporting Information): amino acids are listed
in decreasing order of rates from top to bottom, which does not
match the order of pKa values. Table S2 (Supporting In-
formation): rates are nonidentical, which should not be the case
if pKa was the dominant determinant of the relative rates). When
the order of reactivities is examined by neglecting pH/pKa effects
(Figure 5B and Table S2 (Supporting Information)), the order of
Phe and Ala changes so that the whole order is precisely that
predicted on the basis of nucleophile sterics. With respect to
N-alkylation effects on reactivity, it is clear that the inhibitory
steric effects completely dominate over potentially stimulatory
inductive effects. These conclusions are in line with the chemical
theory that amino nucleophiles are more hindered sterically by
direct substituents than by more distal substituents.
Finally, we consider the reactivities of the other six amino acids
(Ile, Val, Glu, Leu, Lys, and His) shown to have uniform
decoding rates (within a 2-fold error range) on their cognate
tRNAs at the ribosomal A site (1). Of these six amino acids, Lys
and His have side chain nucleophiles that are problematic for our
fMet-NHS assay; therefore, we selected the other four amino
acids to test (Figure 7). The assay was modified to increase
throughput by measuring the relative rates of each individual free
amino acid in competition with an internal standard nucleophile,
N-Me-Gly-NH2. This internal standard was chosen because
of its relatively slow reaction rate and the uncomplicated
NMR spectrum of its product. As a further control, Phe was
also competed with N-Me-Gly-NH2 to aid the comparison of
results in Figures 5A and 8. The measured chemical reactivities
F
IGURE 5: Rates of dipeptide bond formation between fMet-NHS
and amino acid amides in the chemical model of physiological
aqueous pH 7.4. (A) Plots of the apparent rates (min-1) vs concen-
tration of total amine. (B) Plots of apparent rates vs concentration of
unprotonated amine calculated from pKa measurements. Each point
represents a time course at that particular concentration, with each
error bar indicating a goodness-of-fit of each time course curve.
Curvature due to base catalysis is detailed in Materials and Methods
and Tables S1 and S2 (Supporting Information).
ribosomal dipeptide rates are Gly-tRNAGly =Phe-tRNAPhe
=
Ala-tRNAAla > Pro-tRNAPro (1, 3)). The more uniform rates
with cognate tRNAs on the ribosome may be due to tuning by the
tRNAs to decrease differences in the reactivity of the amino acids
(as shown for Pro (3)) and/or a rate-limiting conformational step
for non-Pro-tRNAPro substrates.
We also investigated the chemical principles responsible for the
differences in nonenzymatic rates. In physiological buffer, the
rate of dipeptide formation is expected to be governed by sterics
and also the concentration of unprotonated amine as calculated
by the Henderson-Hasselbalch equation (18). The latter expec-
tation was investigated in our system as follows. First, we
measured the pKa values of the amino acid amides under the
salt and solvent conditions of the dipeptide assay (Figure 3; Table
S3 (Supporting Information)). Then, we found that the change in
aminolysis yield was the same whether it was produced by
changing the concentration of total amine (at constant pH) or
by changing the concentration of free amine (at constant total
amine concentration) through pH changes calculated from our

Article
Biochemistry, Vol. 49, No. 10, 2010 2185
reactivity hypothesis. Evaluation of the effects of N-substitution
sterics, side chain sterics, induction, and pKa in the chemical model
showed that the dominant factor was N-substitution sterics.
ACKNOWLEDGMENT
We thank Donald Stec for help with NMR, Mans Ehrenberg
˚
for discussions and comments on the manuscript, and Brian
Bachmann, Craig Lindsley, and Charles Sanders for comments
on the manuscript.
F
IGURE 7: Amino acid nucleophiles reacted with fMet-NHS in com-
petition reactions with N-Me-glycinamide. The pKa values of the
nucleophiles are Ile, 9.60; Val, 9.52; Glu, 9.58; Phe, 9.09; Leu, 9.58
(CRC Handbook online); and N-Me-Gly-NH2, 8.55 (from Figure 3).
SUPPORTING INFORMATION AVAILABLE
Tables S1-S3 containing kinetic rate constants and pKa values
associated with Figures 3 and 5. This material is available free of
charge via the Internet at http://pubs.acs.org.
REFERENCES
1. Ledoux, S., and Uhlenbeck, O. C. (2008) Different aa-tRNAs are
selected uniformly on the ribosome. Mol. Cell 31, 114–123.
2. Pape, T., Wintermeyer, W., and Rodnina, M. V. (1998) Complete
kinetic mechanism of elongation factor Tu-dependent binding of
aminoacyl-tRNA to the A site of the E. coli ribosome. EMBO J. 17,
7490–7497.
3. Pavlov, M. Y., Watts, R. E., Tan, Z., Cornish, V. W., Ehrenberg, M.,
and Forster, A. C. (2009) Slow peptide bond formation by proline and
other N-alkylamino acids in translation. Proc. Natl. Acad. Sci. U.S.A.
106, 50–54.
4. Zhang, B., Tan, Z., Dickson, L. G., Nalam, M. N., Cornish, V. W.,
and Forster, A. C. (2007) Specificity of translation for N-alkyl amino
acids. J. Am. Chem. Soc. 129, 11316–11317.
F
IGURE 8: Relative reaction rates of six nucleophiles. The rate for the
standard competitor in all assays, N-Me-glycinamide, is normalized
to 1.0, and the rates for the five natural amino acids (free acid forms)
are multiplied by the same factor. Error bars indicate the 95%
confidence interval (2 standard deviations).
5. Weber, A. L., and Miller, S. L. (1981) Reasons for the occurrence of
the twenty coded protein amino acids. J. Mol. Evol. 17, 273–284.
6. Eichler, J., and Houghten, R. A. (1993) Identification of substrate-
analog trypsin inhibitors through the screening of synthetic peptide
combinatorial libraries. Biochemistry 32, 11035–11041.
7. Herman, L. W., Tarr, G., and Kates, S. A. (1996) Optimization of the
synthesis of peptide combinatorial libraries using a one-pot method.
Mol. Diversity 2, 147–155.
8. Kim, H.-S., Chung, T. D., and Kim, H. (2001) Voltammetric deter-
mination of the pKa of various acids in polar aprotic solvents using
1,4-benzoquinone. J. Electroanal. Chem. 498, 209–215.
9. Chambers, R. W., and Carpenter, F. H. (1955) On the preparation
and properties of some amino acid amides. J. Am. Chem. Soc. 77,
1523–1526.
of the natural amino acid nucleophiles are quite similar
(Figure 8), consistent with both the uniform decoding hypothesis
and the chemical reactivity hypothesis. As with the amino acid
amide set in Figure 1, the relative rates do not correlate with the
pKa values of the nucleophiles (pKa values are listed in the
Figure 7 legend, with N-Me-Gly-NH2 and Phe having the lowest
pKa values). Different steric and inductive properties of the side
chains also did not substantially affect reaction rates, at least for
the side chains examined together with the fMet-NHS electro-
phile. However, as with the amino acid amide set in Figure 1,
N-alkylation sterics was again the main rate-determining factor.
In conclusion, dipeptide formation rates on the ribosome with
natural and unnatural amino acid nucleophiles correlate well with
the relative chemical reactivities of the nucleophiles under physio-
logical-like conditions in our chemical model. This correlation
argues against extending the rate-limiting accommodation hy-
pothesis to natural Pro-tRNAPro isoacceptors. Rather, our corre-
lation supports the chemical reactivity hypothesis as the
explanation for slow translation rates with natural Pro-tRNAPro
isoacceptors, dramatically slower incorporations of other N-alkyl
amino acids, and the lack of N-alkyl amino acids other than Pro in
the genetic code. Pro and unnatural amino acids aside, the coup-
ling rates of other amino acids to fMet measured in the chemical
model are similar (as they are on the ribosome), which is compatible
with both the uniform decoding hypothesis and the chemical
10. Schmeing, T. M., Huang, K. S., Kitchen, D. E., Strobel, S. A., and
Steitz, T. A. (2005) Structural insights into the roles of water and the 20
hydroxyl of the P site tRNA in the peptidyl transferase reaction. Mol.
Cell 20, 437–448.
˚
11. Trobro, S., and Aqvist, J. (2005) Mechanism of peptide bond syn-
thesis on the ribosome. Proc. Natl. Acad. Sci. U.S.A. 102, 12395–12400.
12. Okuda, K., Seila, A. C., and Strobel, S. A. (2005) Uncovering the
enzymatic pKa of the ribosomal peptidyl transferase reaction utilizing
a fluorinated puromycin derivative. Biochemistry 44, 6675–6684.
13. Satterthwait, A. C., and Jencks, W. P. (1974) The mechanism of
aminolysis of acetate esters. J. Am. Chem. Soc. 96, 7018–7031.
14. Koshland, D. E. (1951) Kinetics of peptide bond formation. J. Am.
Chem. Soc. 73, 4103–4108.
15. Notari, R. E. (1969) Hydroxamic acids II: Kinetics and mechanisms of
hydroxyaminolysis of succinimide. J. Pharm. Sci. 58, 1064–1068.
16. Salomaa, P., Schaleger, L. L., and Long, F. A. (1964) Solvent
deuterium (D) isotope effects on acid-base equilibriums. J. Am. Chem.
Soc. 86, 1–7.
17. Krezel, A., and Bal, W. (2004) A formula for correlating pKa values
determined in D2O and H2O. J. Inorg. Biochem. 98, 161–166.
18. Schroeder, G. K., and Wolfenden, R. (2007) The rate enhancement
produced by the ribosome: an improved model. Biochemistry 46,
4037–4044.