Process improvements for the manufacture of tenofovir disoproxil fumarate at commercial scale

Article abstract of DOI:10.1021/op1001337

The three-step manufacturing process used in the synthesis of tenofovir disoproxil fumarate (1) was studied and optimized, leading to a more productive and robust process. The yield was improved from about 13% overall to 24%. Key process improvements identified included implementation of a telescoped process for the second stage that obviated the need for an extraction and solvent exchange, and significant optimization of the final reaction, including the beneficial effect of adding a quaternary ammonium salt to the alkylation reaction and development of a nonaqueous process for removal of NMP and triethylamine from the product mixture to decrease the level of decomposition of product during the isolation.

Full text of DOI:10.1021/op1001337

Organic Process Research & Development 2010, 14, 1194–1201
Process Improvements for the Manufacture of Tenofovir Disoproxil Fumarate at
Commercial Scale
David H. Brown Ripin,* David S. Teager,* Joseph Fortunak,† Shaik Mahaboob Basha,| Nylea Bivins,† Christopher N. Boddy,‡
Stephen Byrn,§ Kelly K. Catlin, Stephen R. Houghton,‡ S. Tirumala Jagadeesh,9 K. Anesh Kumar,9 Jack Melton,⊥
Shaik Muneer,| L. Nagaprasada Rao,9 R. Venkateswara Rao,9 Puma Chandra Ray,9 Nardla Gopal Reddy,|
Ravi Mallikarjuna Reddy,| K. Chandra Shekar,9 Tricia Silverton,† Daniel T. Smith,§ Rodger W. Stringham,
Gottumukkala V. Subbaraju,| Frajovon Talley,† and Adrian Williams†
Clinton Health Access InitiatiVe, 383 Dorchester AVenue, Suite 400, Boston, Massachusetts 02127, U.S.A., Department of
Chemistry, Howard UniVersity, 525 College Street NW, Washington, D.C. 20059, U.S.A., Department of Chemistry, UniVersity
of Ottawa, Ottawa, Ontario ON K1N 6N5, Canada, Department of Industrial and Physical Pharmacy, Purdue UniVersity, 575
Stadium Mall DriVe, West Lafayette, Indiana 47907-2091, U.S.A., ProcesScience, Inc., 117 Woodberry Lane, FayetteVille,
New York 13066, U.S.A., Inogent Laboratories, IDA Nacharam, Hyderabad - 500 076, India, and Aptuit Laurus PVt. Ltd.,
ICICI Knowledge Park, Turkapally, Hyderabad - 500 078, India
Scheme 1. Tenofovir disoproxil fumarate and its active
metabolite, tenofovir
Abstract:
The three-step manufacturing process used in the synthesis of
tenofovir disoproxil fumarate (1) was studied and optimized,
leading to a more productive and robust process. The yield was
improved from about 13% overall to 24%. Key process improve-
ments identified included implementation of a telescoped process
for the second stage that obviated the need for an extraction and
solvent exchange, and significant optimization of the final reaction,
including the beneficial effect of adding a quaternary ammonium
salt to the alkylation reaction and development of a nonaqueous
process for removal of NMP and triethylamine from the product
mixture to decrease the level of decomposition of product during
the isolation.
From a manufacturing perspective, the diminished stability
in water that makes 1 an effective pro-drug also adds to the
difficulty in manufacturing and isolating this product. Hydrolysis
of the disoproxil side chains occurs readily under a range of
conditions, complicating the workup and isolation of the
tenofovir disoproxil free base. As a result, the design of the
reaction and isolation to form the API represents the most
critical portion of the manufacturing process.
TDF (1) is manufactured in a three-stage, four-reaction
process as shown in Scheme 2. Reaction of readily available
adenine (3) with R-propylene carbonate (4) is followed by base-
promoted alkylation of the secondary alcohol with tosylated
hydroxymethylphosponate diester, 6. Hydrolysis of the diethyl
phosphonate esters produces tenofovir (2). An alkylative
esterification completes the process to produce crude free
base, 9, which is then treated with fumaric acid to
crystallize TDF (1).
Introduction
Tenofovir disoproxil fumarate (1, or TDF) is a nucleoside
analogue reverse transcriptase inhibitor (nRTI) that is used in
the treatment of HIV/AIDS and hepatitis B.1 TDF is a pro-
drug of tenofovir (2, also known as PMPA) designed to improve
absorption and cell permeability of the active moiety. See
Scheme 1. There are a number of advantages to the use of
tenofovir in HIV therapy over the alternatives zidovudine (AZT)
and stavudine (d4T), including a long half-life (allowing for
once-daily dosing), a lower rate of resistance development, and
improved side-effect profiles when compared to the older
medications.2
The yields for this process3 are fair to modest, with an overall
yield from 3 of about 13%. The third stage, for the reason
described above, is particularly challenging, with isolated yields
of 1 of only about 35% based upon 2. This contribution reports
improvements throughout the process, with a particular focus
on the stage 3 chemical transformation and isolation procedures.
* Authors to whom correspondence should be addressed: E-mail: dripin@
clintonhealthaccess.org; dteager@clintonhealthaccess.org.
† Department of Chemistry, Howard University.
‡ Department of Chemistry, University of Ottawa.
§ Department of Industrial and Physical Pharmacy, Purdue University.
⊥ ProcesScience, Inc.
9 Inogent Laboratories.
| Aptuit Laurus Pvt. Ltd.
Results and Discussion
(1) (a) Arimilli, M. N.; Kim, C. U.; Dougherty, J.; Mulato, A.; Oliyai,
R.; Shaw, J. P.; Cundy, K. C.; Bischofberger, N. AntiViral Chem.
Chemother. 1997, 8, 557. (b) Naessens, L.; Bischofberger, N.;
Augustijns, P.; Annaert, P.; Van Den Mooter, G.; Arimilli, M. N.;
Kim, C. U.; De Clerq, E. Antimicrob. Agents Chemother. 1998, 42,
1568. (c) Robbins, B. L.; Srinivas, R. V.; Kim, C.; Bischofberger, N.;
Fridland, A. Antimicrob. Agents Chemother. 1998, 42, 612.
(2) Balzarini, J.; Holy, A.; Jindrich, J.; Naesens, L.; Snoeck, R.; Schols,
D.; De Clerq, E. Antimicrob. Agents Chemother. 1993, 37, 332.
Stage 1. Synthesis of Hydroxypropyl-adenine (5). Adenine
(3) and R-propylene carbonate (4) react readily at 120 °C in
the presence of sodium hydroxide in DMF to form alcohol 5
in high yield. R-propylene carbonate (4) is used in excess (1.3
(3) Arimilli, M. N.; Cundy, K. C.; Dougherty, J. P.; Kim, C. U.; Oliyai,
R.; Stella, V. J. U.S. Patent 5,922,695, 1998.
1194
•
Vol. 14, No. 5, 2010 / Organic Process Research & Development
10.1021/op1001337  2010 American Chemical Society
Published on Web 08/24/2010

Scheme 2. Manufacturing route to tenofovir disoproxil fumarate
Table 1. Selection of base for stage 2aa
base conversion to 7 (%)
equiv) to ensure that the reaction goes to completion; the
reaction proceeds relatively quickly, while 4 and the desired
product 5 were confirmed to be quite stable under the reaction
conditions. The only significant impurity formed is a regioiso-
mer, assigned structure 10 (Figure 1), in an approximately 10:1
product:impurity ratio. Precipitation of solids from the
reaction mixture by addition of toluene provides an 84%
yield of crude 5, contaminated with 7-8 area % of 10.
However, we found that the product can be crystallized
in 66% yield from the DMF reaction solvent by addition
of a 1:1 MeOH/i-PrOH mixture to provide product of 97.8
area % purity, with only 1.7% regioisomer remaining. This
impurity can be carried forward in the higher amount to
the next stage of the process, but we elected to reduce
the impurity level at this point in the process, in order to
maximize utilization of reagents and reduce the amount
of potential impurities in the second reaction stage.
Stage 2. Preparation of Phosphonic Acid 2. The alcohol
5 is converted into 2 in a two-reaction sequence involving
alkylation with the tosylate 6 followed by hydrolysis. The two
reactions are optimally telescoped together, as the intermediate
product, the diethylphosphonate, 7, is water-soluble, prone to
hydrolysis to the corresponding monoester, and difficult to
crystallize. The alkylation reaction can be run in a number of
polar aprotic solvents, with DMF and NMP giving best results.
The presence of DMF in the second stage of the sequence (de-
esterification reaction) leads to some dimethylphosphamide
impurity formation as a result of the presence of free dimethyl-
amine as a contaminant in DMF, leading to the selection of
NMP for the two-stage process.
LiO-t-Bu
61
35
59
39
0
LiO-i-Pr
LiO-CH2C(CH3)3
NaO-t-Bu
NaO-i-Pr
NaO-CH2CH(Et)C4H9
KO-t-Bu
Mg(O-t-Bu)2(MTB)
LiHMDS
68
10
90
52
a Standard conditions: 5 (1 equiv), base (3 equiv), NMP (5 vol), 6 (1.5 equiv),
70 °C.
Stage 2a. The alkylation can, in principle, be effected with
a number of bases other than the reported lithium tert-butoxide3,4
and magnesium tert-butoxide (MTB).5 Our trials with various
hindered alkoxide bases and with lithium hexamethyldisilazide
(Table 1) show that magnesium tert-butoxide (MTB) provides
a significantly higher in situ conversion to 7 than any alternative
investigated. Other bases including carbonates, DBU, tetram-
ethylguanidine, and phosphazene bases provided less than 50%
conversion.
We noted, however, variation in results dependent upon the
batch or supplier of MTB. A number of additives (potential
ionic contaminants) were studied in this reaction in an effort to
determine what might affect the performance of the MTB (Table
2), and analytical tests (including base titration, KF, IR, TGA,
and a determination of tert-butyl alcohol levels) were employed.
While some correlations were observed, none of these tests
proved predictive of reaction success, although control of halide
content seems the most likely factor to consider. The authors
recommend use-testing of all material to ensure successful
reactions.
Stage 2b. The hydrolysis of diester 7 to tenofovir (2)
proceeds via a Lewis acid-promoted nucleophilic dealkylation.6
The previous reported conditions for this reaction are the use
of a substantial excess (6 equiv) of TMSBr. In addition to being
(4) Schultze, L. M.; Chapman, H. H.; Dubree, N. J. P.; Jones, R. J.; Kent,
K. M.; Lee, T. T.; Louie, M. S.; Postich, M. J.; Prisbe, E. J.; Rohloff,
J. C.; Yu, R. H. Tetrahedron Lett. 1998, 39, 1853.
(5) Becker, M. W.; Chapman, H. H.; Cihlar, T.; Eisenberg, E. J.; He,
G.-X.; Kernan, M. R.; Lee, W. A.; Prisbe, E. J.; Rohloff, J. C.;
Sparacino, M. L. PCT Intl. WO/0208241A32002.
(6) Houghton, S. R.; Melton, J.; Fortunak, J.; Brown Ripin, D. H.; Boddy,
C. N. Manuscript in preparation.
Figure 1. Stage 1 key impurity.
Vol. 14, No. 5, 2010 / Organic Process Research & Development
•
1195

Table 2. Effect of additives on conversion in 5 + 6 with
MTB
Scheme 3. Key impurities formed in the stage 3a process
conversion
with 0.1
equiv (%)
conversion
with 0.5
equiv (%)
conversion
with 1.0
equiv (%)
additive
MgBr2
MgCl2
Mg(OH)2
NaBr
65
82
84
69
86
84
88
33
80
83
59
85
62
85
66
30
84
38
79
46
82
NaCl
NaOH
NaOTs
expensive and variable in quality, TMSBr requires special
handling to minimize the destructive impact of this reagent on
the materials of reactor construction. Two practical alternatives
to TMSBr for the execution of this process are available. Either
aqueous HBr7 or the use of TMSCl in the presence of NaBr in
organic solvent6 was preferred for process stage 2b. The use of
aqueous HBr or other mineral acid for the de-esterification
requires first a high-vacuum distillation to remove organic
solvent (e.g., DMF) or a high-volume dichloromethane (DCM)
extraction of the highly water-soluble intermediate (7) from the
water/NMP (or DMF) layer following the alkylation reaction
(stage 2a).
As a result, we chose to optimize a telescoped process
wherein the alkylation reaction is run in NMP, taken through
nonaqueous workup to remove salts, and hydrolyzed in the
remaining NMP using TMSCl and NaBr. Workup and pH
adjustment precipitates phosphonic acid 2 as its hydrate. This
process precludes the need for an intermediate product extraction.
The mass balance of the process was monitored to determine
where product loss occurs. The alkylation with 6 reproducibly
provides an 85-90% in situ yield of the intermediate product
7, with the remainder of the mass being residual starting material
and a number of lower-level impurities. In the magnesium salt
cake, a 15% loss of 7 is observed; some of this loss can be
recovered in a manufacturing setting with a stirred filter but
can be difficult in a lab setting where the salt cake quickly
absorbs atmospheric water. This intermediate is hydrolyzed to
2 via the intermediate monoethyl phosphonate ester in 95%
conversion. Following the addition of water, little product is
lost in the subsequent organic extractions to remove residual
organic material, and a 55% yield of product 2 (from 5) was
isolated in the experimental run, with 13% of product remaining
in the aqueous mother liquors.
an aqueous component at pH 2.5 was ideal. Using these
conditions the major impurities are well resolved. In the
chromatogram of a reaction (not optimized) shown in Figure
2, 2 elutes at a retention time of 0.82 min followed by 11 and
14 at 0.97 and 1.17 min. The diester 9 elutes at 5.60 min, well
separated from the main N-hydroxylmethyl impurity 12 at 6.26
min. It was observed during development that use of a higher
pH buffer results in less effective separation of impurities and
column performance that deteriorates rapidly over multiple
injections. At pH 5.5 no resolution is observed. In addition to
its use in reaction monitoring, this method has been used
successfully for assaying process streams as well as evaluation
of isolated API. This method allowed us to improve the
throughput, reproducibility, and scalability of the stage 3
process.
Stage 3. Research Objectives. Stage 3 of the TDF process
comprises three operations. The first is the reaction of 2 to yield
diester 9 in solution. The second is the workup of diester 9 to
partially remove impurities and provide 9 suitable for conversion
to the fumarate salt 1. The third operation is formation of 1
itself. This discussion primarily focuses on optimization of the
first and second of these operations.
Stage 3a: Synthesis of 9. The conversion of 2 into 9 by base-
promoted alkylation with excess chloromethyl isopropyl car-
bonate (8) is accompanied by physical realities that complicate
the chemical reaction and workup. Due to the poor solubility
of 2 and its salts, the stage 3a reaction must be run in a polar,
aprotic solvent, with non-nucleophilic amine bases being
preferred (triethylamine, DIPEA, TMG, and 2,6-lutidine are
notable). The phosphonic acid 2 is not initially soluble in NMP
(or other solvents such as DMF, DMSO, or acetonitrile).
Addition of triethylamine to a stirred suspension of 2 in NMP
at 40-60 °C causes the rapid dissolution of 2 to give a solution.
After a short period of stirring, however, a thick suspension is
formed by precipitation of a mixture of mono- and bis-
triethylammonium salts of 2. Upon addition of 8, alkylative
esterification gradually occurs; monoester 11 and diester 9 are
soluble in the reaction medium, and the suspension of salts gives
way to a solution as the reaction proceeds. The conversion of
11 to 9 is significantly slower than the initial alkylation of 2 to
give 11. As conversion to 9 proceeds, the reverse reaction (de-
esterification of 9 to 11) becomes competitive with the rate of
the forward reaction itself. We have confirmed that this is purely
an effect of the slow conversion of 11 to 9, while reagent 8
Stage 3a Process Analytical Technology. Assay for
Formation of Tenofovir Disoproxil (9). Before work could
be done to optimize the formation and isolation of 1 from 2,
an effective analytical method for the assay of the stage 3a
process had to be developed. Several key impurities (including
persistent reaction intermediate 11 along with N-hydroxy-
methylated impurities 12 and 14, and isopropylcarbamate
impurities 13 and 15) are consistently observed in the stage 3a
process (Scheme 3).
The ability to separate the N-hydroxymethyl impurities
proved to be highly dependent on the pH of the mobile phase;
(7) Vasireddy, U. M. R.; Vellanki, S. R. P.; Balusu, R. B.; Bandi, N. D. R.;
Jujjavarapu, P. K.; Ginjupalli, S. R.; Pilli, R. K. PCT Intl. WO2008/
007392A2, 2008.
1196
•
Vol. 14, No. 5, 2010 / Organic Process Research & Development

Figure 2. HPLC of reaction mixture with pH 2.5 aqueous eluent.
Table 3. Effect of solvent volume on stage 3a conversiona
and triethylamine do not react together or otherwise undergo
significant decomposition under the reaction conditions.
Running the reaction using the conditions published by
Gilead3 (4 vols of NMP, 4 equiv of TEA, 4.7 equiv of 8,8 60
°C, 4 h), a typical reaction result is approximately 50-65% in
situ yield, with 5-10% of the N-hydroxymethylated impurity
12 and 10-15% of the monoester intermediate 11 (and 2-5%
its N-hydroxymethylated analogue 14) present along with other
lower level impurities, including 13 and 15. Efforts to push the
reaction further to completion lead to product decomposition
at a greater rate than product formation. Indeed, the product
quality degrades steadily once the reaction reaches maximum
conversion, leading inevitably to lower yields as the process is
scaled (due to increased process time with scale). Isolated yields
for the reported procedure rarely exceed 35%.
time to max
volumes product monoester
N-CH2OH
in situ
NMP (9) (%) (11) (%) impurity (12) (%) yield of 9 (h)
1.5
3
58.3
63.7
63.3
64.0
67.1
69.9
7.4
18.6
12.1
9.9
10.1
7.7
11.5
6
6
7
7
9
3
10
4
15.3
12.4
13.2
6.3
6
8
8b
a Standard conditions: 5 equiv of 8, 4 equiv of TEA base, 50° C. b Reaction run
with 8 equiv of 8.
and the relative success of the reaction was based on the
maximum conversion (area % of 9 by HPLC) found. As well,
area percent levels of key intermediates and process impurities
were also monitored, and are reported below.
The isolation of product also proved to be problematic, with
significant degradation observed throughout the procedure.
Product decomposition during workup is primarily a result of
aqueous hydrolysis, with NMP bringing water into the organic
layer of the extractions and taking product into the aqueous
layer. Additionally, the presence of TEA in the workup leads
to a basic water layer (pH ) 10), promoting this hydrolysis.
With several back-extractions required to maximize recovery
of product, the process proved to be extremely sensitive to scale.
Therefore, an intensive investigation was undertaken to
improve conversion, product stability during isolation, and
overall robustness of the process. Several critical processing
parameters were identified that affect the outcome of both the
reaction to make 9 and its isolation. These include base
selection, reaction concentration, water content in the starting
material and reaction mixture, and reaction temperature. As the
starting material mixture is sparingly soluble in the reaction
solvent, addition of phase-transfer reagents was also investi-
gated. Once the above parameters were largely optimized, a
follow-up investigation considered the reaction solvent chosen,
in the event that the improved conditions might allow a solvent
choice making for easier product isolation. For the workup and
isolation, solvent considerations (e.g., extractions), pH, and
processing time were seen as critical to maximizing quality
and throughput.
Optimization of Formation and Isolation of TenofoVir
Disoproxil (9). Base. Bases screened included inorganic bases
Na2CO3, K2CO3, and Li2CO3, and organic bases triethylamine
(TEA), N,N-diisopropylethylamine, tetramethylguanidine, and
2,6-lutidine. Inorganic bases produced insoluble phosphonic acid
salts and negligible conversions. Among organic bases, TEA
provided the best performance in the reaction. Use of three or
more equivalents provided optimum conversion. Running the
reaction with prepared lithium, sodium, or potassium salts of 2
in the presence of TEA also led to low conversions, presumably
due to the poor solubility of the salt.
Reaction Concentration. Selecting an optimal reaction
concentration (solvent volume) is a balancing act, as illustrated
by the results in Table 3: more N-hydroxymethylated impurity
is formed at higher concentrations, therefore favoring more
solvent/higher dilution. However, there is a practical upper limit
to the amount of NMP that can be used, with respect to the
aqueous workup, that favors lower solvent amounts. At higher
dilution the reaction is enhanced by additional quantities of 8,
but at significant additional cost. These experiments led to
selection of four volumes of NMP as the optimum; this provides
a relatively thick, but stirrable, reaction mixture.
Water. Water is a major factor in impurity formation and in
product stability. When water is present in the stage 3a reaction,
formation of the N-hydroxymethylated impurities 12 and 14 is
enhanced. If the stable, hydrated, form of 9 is used directly,
resulting in water content of >6000 ppm in the reaction mixture,
these impurities may comprise >25% of the product mixture.
With reaction mixtures at <500 ppm water, 5-10% impurity
For comparison of reaction conditions, reaction mixtures
were sampled during the alkylation stage at various time points,
(8) For studies in this paper, 5.0 equiv of 8 were used. Further optimization
is possible.
Vol. 14, No. 5, 2010 / Organic Process Research & Development
•
1197

Figure 3. Effect of water on stage 3a: hydrated vs dehydrated 2. Standard conditions: 4 volumes NMP, 5 equiv 8, 4 equiv
TEA, 50 °C.
formation is observed (see Figure 3). There are a number of
possible mechanisms for the formation of these impurities, and
the fact that there is still N-hydroxymethylated byproduct
formation even in the driest reactions (<50 ppm water) seems
to indicate that more than one mechanism for the formation of
formaldehyde or a reactive equivalent are at work. Lower
amounts of hydroxymethlyated impurity are formed at higher
dilution, supporting this hypothesis.
Two methods for controlling water levels were identified:
drying 2 at high temperature to drive off crystalline water, and
using a solvent for azeotropic drying of the suspension of 2 in
NMP prior to addition of reagents in the stage 3a process. The
optimal and most efficient process for drying may be to
dehydrate 2 at 70-90 °C under vacuum in an agitated dryer,
to provide the anhydrous form of 2 prior to the subsequent
reaction. Anhydrous 2 is hygroscopic, forming again the stable
monohydrate, and care must be taken to prevent any rehydration
prior to the stage 3a process. Alternatively, azeotropic distillation
of a suspension of 2 in NMP using either cyclohexane or toluene
is effective for producing a reaction mixture of low water
content. In practice, in order to ensure low water content, both
high-temperature drying of the solid and azeotropic drying of
the NMP suspension were used for scale-up studies.
heterogeneous until conversion nears its maximum, it was
postulated that a phase-transfer catalyst might enhance the rate
and conversion in the reaction. Additionally, the in situ
generation of a more reactive derivative of 8, such as the
bromide or iodide, might also improve the reaction profile. To
test this, a number of additives were tested in the reaction (Table
4). A significant enhancement of the reaction rate was observed
when tetrabutylammonium bromide (TBAB) was included, with
the reaction reaching maximum conversion at 4-6 h with
additive vs 8-10 h without (Figure 4). The lack of improvement
on addition of Bu4NCl and Bu4NOTs indicates that the
improvements are likely a result of the presence of bromide,
with a likely positive effect due to enhanced solubility of the
salts of 2. However, we note that the bromo and iodo analogues
of 8swhich may be prepared but are unstable to isolation as
neat substancessdo not give better reactions than does 8 itself.
There is some enhancement at substoichiometric levels of
additive, but the best results were obtained with one equivalent
of tetrabutylammonium bromide. Other tetraalkylammonium
bromides may also be suitable; the reaction was comparably
successful with BnBu3NBr.
As shown in Table 4, the use of a phase transfer reagent
not only improved the reaction rate but also appeared to stabilize
the product in the reaction mixture, maintaining a low rate of
formation of impurities (such as 12, cf. Figures 3 and 4) under
AdditiVe. As the TEA salt of 2 (as initially formed in the
reaction) has limited solubility in NMP, and the reaction is
Table 4. Effect of phase transfer reagent additives on stage 3a conversiona
additive equiv of additive product (9) (%) monoester (11) (%) N-CH2OH (12) (%) time to max in situ yield of 9 (h)
none
0
63.3
64.8
74.5
75.6
71.5
67.6
38.9
72.0
66.3
0
15.3
15.3
4.8
9.9
12.0
10.0
10.5
8.8
11.4
12.5
9.4
10 h
8 h
3 h
4 h
6 h
7 h
8 h
6 h
8 h
n/a
Bu4NCl
1
Bu4NBr
1.5
Bu4NBr
1
5.8
Bu4NBr
0.5
8.6
Bu4NBr
0.1
11.9
13.7
9.4
Bu4NBr + NaI
BnBu3NBr
Bu4NOTs
NaOTs
0.5 each
1
1
3
9.3
7.0
13.7
0
a Standard conditions: 4 volumes NMP vs 2, 5 equiv 8, 4 equiv of TEA base, 50 °C.
1198
•
Vol. 14, No. 5, 2010 / Organic Process Research & Development

Figure 4. Reaction profile with and without added TBAB.
Figure 5. Temperature optimization, with phase transfer reagent. Standard conditions: 4 volumes NMP, 5 equiv 8, 4 equiv TEA,
1 equiv TBAB.
the reaction conditions. In a separate experiment, when 9 is
subjected to the reaction conditions, the reverse reaction
(hydrolysis) as well as formation of impurities 12 and 14 is
slower in the presence of TBAB than in its absence.
Table 5. Stage 3a conversion in different solventsa
time to max
product monoester N-CH2OH in situ yield
solvent
(9) (%)
(11) (%)
(12) (%)
of 9 (h)
Temperature. The combination of water removal and addi-
tion of TBAB allowed for the reaction temperature to be
lowered from the reported reaction temperature of 65-70 °C.
As can be seen from Figure 5, reducing the reaction temperature
from 70 to 60 °C or lower leads to a significant improvement
in the stability of the product under the reaction conditions.
When the phase transfer reagent is used, a temperature range
of 50-60 °C is indicated, maximizing conversion, stability, and
reduced reaction time.
SolVent. A polar aprotic solvent is necessary for the alkyl-
ation reaction to proceed adequately, primarily driven by
solubilities of the starting material, its salts, and salts of the
monoalkylated intermediate.9 The reported best solvent for the
reaction is NMP, but workup of reactions using this solvent
can be more difficult when compared with other polar aprotic
solvent options. Therefore, once the process improvements
ethyl acetate
acetonitrile
DMSO
14.9
48
52.5
63.6
75.6
12.0
17.5
13.9
11.2
5.8
7.4
10.2
13.1
14.5
10.5
9 h
8 h
3 h
4 h
4 h
DMF
NMP
a Standard conditions: 5 volumes solvent, 5 equiv 8, 1 equiv TBAB, 4 equiv
TEA, 50 °C.
above had been identified, we undertook a cursory examination
of solvent selection using the improved conditions (Table 5).
The reaction proceeds at a reasonable rate in acetonitrile, but
the formation of new impurities is observed, presumably from
reaction of acetonitrile with 8, followed by reaction of the
(9) Poor sampling technique can lead to the impression that the reaction
is proceeding cleanly in less polar solvents, while in actuality the
conversion is low.
Vol. 14, No. 5, 2010 / Organic Process Research & Development
•
1199

activated intermediate (based on MS data). In DMF, higher
levels of N-hydroxymethylated impurities as well as solvent-
derived N,N-dimethylphosphamide impurities are observed. The
use of DMSO leads to significant impurity formation, including
oxidation byproduct. As well, the reverse reaction rate of 9 to
2 is enhanced in DMF and DMSO relative to NMP. Thus, NMP
was confirmed as the optimal solvent for the stage 3a process.
Stage 3. Improved Workup of 9 and Isolation of 1. Once
the stage 3a reaction is completed, a significant amount of work
is required to isolate 9 of sufficient purity to form 1 (TDF) that
meets specifications. The published procedure3 includes pre-
cipitation of triethylamine salts by addition of isopropyl acetate,
and aqueous workup of the filtered solution to remove the NMP
and additional salts from the solution of 9. It was noted that
the pH of the aqueous phase during this workup was ap-
proximately 10. Under these conditions, decomposition of the
product was measured as the workup progressed. As the reaction
process uses excess amine base, which would cause this elevated
pH on workup, we examined the removal of excess triethyl-
amine by concentration of the reaction mixture, after addition
of the ethyl acetate to precipitate salts. Co-distillation of
triethylamine with ethyl acetate provided a mixture that (without
filtration of salts) gave an aqueous pH of 4 during workup, with
notably lower levels of decomposition products. In an early
comparison trial (prior to other optimizations noted above), the
concentration process provided a 43% product yield compared
with 35% for the filtration process. However, the codistillation
process extended the process times, so an alternative procedure
was sought.
After significant experimentation, it was found that addition
of cyclohexane at the end of the reaction results in a two-phase
mixture, provided that four or fewer volumes of NMP are used
in the reaction. Under these conditions, only trace amounts of
9 transfer into the cyclohexane layer, but excess TEA and a
significant amount of NMP and reagent 8 partition into the
cyclohexane. Three extractions with cyclohexane (5-6 volumes
each) were found to provide an optimal balance between NMP
removal and ability to manipulate the product-containing
reaction phase, which becomes rather thick. The fluid cyclo-
hexane layers must be removed from the top of the thick
reaction mass, requiring the use of a dip-tube or other decanting
technique. In this way, 9 is concentrated in a decreasing amount
of NMP, and excess reagents and some byproduct are re-
moved.10
Conclusion
Several improvements were made to the tenofovir manu-
facturing process. Key findings included the use of a telescoped
procedure for the stage 2 process, identifying high-quality
magnesium tert-butoxide as the optimum base for that process,
and utilizing the lower-cost TMSCl/NaBr reagent mixture in
place of TMSBr. In the final stage of the process, the
introduction of tetrabutylammonium bromide significantly
improved the reaction profile, and the development of a
cyclohexane extraction to remove excess reagents in the absence
of water significantly reduced the level of product decomposi-
tion during the workup and isolation. As a result of these
changes, the overall process yield was improved to 24.3% from
the reported 13% reported by published procedures, resulting
in a significant reduction in the cost of treating patients with
HIV/AIDS.
Experimental Section
Unless noted otherwise, purities are reported as HPLC A%
results with detection at 260 nm. HPLC conditions for stage 1:
Zorbax XDB-C18 or equivalent column (15 cm ×0.46 cm, 3.5
µm particles), isocratic at 1.0 mL/min, mobile phase 900 mL
of 10 mM phosphate buffer adjusted to pH 5.5 mixed with 100
mL 1:1 acetonitrile/methanol. HPLC conditions for stage 2:
same as stage 1, but using the mobile phase above mixed 89:
11 with 1:1 acetonitrile/methanol (to give a net 20% organic
mobile phase). HPLC conditions for stage 3: Zorbax RX-C18
column (15 cm ×0.46 cm, 3.5 µm particles), linear 15-min
gradient of 15% to 70% acetonitrile vs a pH 2.5 10 mM
phosphate buffer, 1.5 mL/min flow, 40 °C.
Preparation of (R)-9-(2-Hydroxypropyl)adenine (5). Ad-
enine (3, 20.0 g, 0.148 mol) and NaOH (0.47 g, 0.08 equiv)
were mixed with DMF (95 mL) at 25-30 °C, and the mixture
was stirred for 10 min at 25-30 °C. R-Propylene carbonate (4,
19.6 g, 1.32 equiv) was added to the reaction mass over 10-15
min at 25-30 °C. The mixture was heated to 120 °C and held
for 24 h. Conversion, monitored by HPLC, reached about 90%
after about 22 h (with regioisomer content about 8%).
The reaction mass was cooled to 70 °C, and a mixture of
methanol (60 mL) and 2-propanol (60 mL) was added over
15-20 min. During addition of solvent, the reaction mass
temperature was allowed to cool to about 55 °C, and precipita-
tion was observed. The reaction mass was cooled to 15 °C and
held at this temperature for about 1 h. The product was isolated
by filtration and the cake washed with a mixture of methanol
(10 mL) and 2-propanol (10 mL) chilled to 5 °C. The solids
were dried under vacuum at 70-75 °C, providing 18.8 g
(65.9%) of alcohol 5. HPLC purity of 5: 97.8% with regioi-
somer 9, 1.7%. Note that 9 has a significantly different UV
spectrum from 3 and 5, so that A% measurements may not
accurately reflect mass ratios of 5 and 9.
Partitioning the residue between ethyl acetate and water
provides a solution of crude 9. Notably, the aqueous pH during
these extractions is 5-6, showing that excess TEA has been
removed. The product 9 exhibits much greater stability in both
the aqueous and organic phases throughout the extraction
process, a vital consideration given the expected extended
processing times on scale-up. Following solvent exchange to
isopropanol, 1 is crystallized by addition of fumaric acid and
isolated in acceptable purity. Using this optimized process in
comparison to the published procedure, recovered yields for
stage 3 improve from about 35% to 60-65%.
Preparation of (R)-(((Adenin-9-yl)propan-2-oxy)meth-
yl)phosphonic Acid (7). Alcohol 5 (4.00 g, 0.021 mol) was
mixed with NMP (20.0 mL) and magnesium tert-butoxide (10.6
g, 3 equiv) at 25-30 °C and heated to 70 °C. Tosylate 6 (10.0
g, 1.5 equiv) was added to the stirred mixture at 70-74 °C
over 10-15 min. Progress of the reaction was monitored by
HPLC; conversion reached 90% at 7 h.
(10) In one experiment, the cyclohexane extracts were concentrated in
vacuo, and the resulting mixture, containing 8, was recycled for use
in the stage 3 process. Provided that moisture is excluded throughout
the process, this recycling process works efficiently.
1200
•
Vol. 14, No. 5, 2010 / Organic Process Research & Development

The mixture was cooled to 15-20 °C. The pH was adjusted
to 6-7 with acetic acid (∼8.0 mL) at 15-25 °C, ethyl acetate
(100 mL) was added with vigorous stirring to precipitate
magnesium salts, and the mixture was heated to 50-60 °C.
The mixture was stirred for 30 min at 50-60 °C and cooled,
and the precipitated salts were filtered at 45-47 °C. Note that
the salts are very hygroscopic, becoming difficult to filter on
exposure to a moist atmosphere; at laboratory scale it was more
practical to decant the liquors from the solid mass. To extract
additional product, the salt cake was taken up in additional ethyl
acetate (75 mL) and stirred for 30 min at 55-60 °C, then filtered
at 45-47 °C. The salts were washed with ethyl acetate (25 mL),
and all filtrates were combined and concentrated on a rotary
evaporator under reduced pressure below 55 °C.11 Care should
be taken during this process to avoid taking up adventitious
moisture. The moisture content of the final concentrate should
be below approximately 0.9% to proceed to the next stage. An
azeotropic drying process using cyclohexane was used in the
laboratory to ensure dryness, chasing the ethyl acetate with one
or two distillations of cyclohexane (30 mL) as needed. In a
manufacturing environment, this should not be necessary.
The solution above was transferred to a reaction flask, and
sodium bromide (7.5 g, 3.5 equiv based on 5) was added. The
mixture was cooled to 0-5 °C. Trimethylsilyl chloride (12.0
g, 5.3 equiv based on 5) was added over 10 min, and the
reaction mixture was heated to 75 °C. Progress of the reaction
was monitored by HPLC; the reaction was considered complete
(>90% of 2) at 16 h.
°C, and triethylamine (70.4 g, 4.0 equiv) was added, causing
precipitation of triethylamine salts of 2. Tetrabutylammonium
bromide (56.1 g, 1.0 equiv, water content <1%) was added,
and the resulting mixture was heated to 50 °C. Chloromethyl
isopropyl carbonate (8, 132.8 g, 5.0 equiv) was added at this
temperature, and stirring continued. The reaction was monitored
for completion; the reaction reached 75 area % of 9 at 5.5 h, at
which point the originally thick suspension was nearly clear.
The mass was cooled to 25-30 °C, and cyclohexane (300 mL)
was added. After the mixture stirred for 15 min and settled for
30 min, the upper phase was removed by suction, providing
280 mL. Cyclohexane (200 mL) was again added and the
process repeated, providing an upper layer of 260 mL. Analysis
of the cyclohexane washes showed no product. To the remaining
reaction mixture were added ethyl acetate (500 mL) and water
(250 mL), and the mixture was stirred for 15 min and settled
for separation. The aqueous phase from the separation was
extracted twice more with ethyl acetate (150 mL each). The
combined organic phases were washed three times with chilled
(5 °C) water (300 mL each) and once with 10% aqueous NaCl.
The resulting organic phase was dried with solid sodium sulfate
and concentrated under reduced pressure. Cooling the residue
to about 5 °C provided a waxy, solid mass of 90.6 g (100%
yield if pure; HPLC purity 84.3 A%).
This crude 9 (90.0 g) was taken up in 2-propanol (315 mL)
at ambient temperature. To this solution was added fumaric acid
(26.1 g, 1.3 equiv), and the mixture was heated to 50 °C. The
resulting solution was stirred at this temperature for 2 h, then
cooled in stages to 5 °C, and stirred an additional hour to fully
crystallize the product. The resulting crystals were isolated by
filtration, washing with a mixture of 2-propanol (75 mL) and
cyclohexane (150 mL) chilled to 5 °C. The solids were dried
under vacuum at 35 °C to provide 69.0 g of 1 (TDF, 62.3%
yield, HPLC purity 98.4%).
The reaction mixture was cooled to 20-25 °C, diluted with
water (40 mL) and washed twice by extraction with ethyl acetate
(2 × 40 mL). The aqueous layer was cooled to about 5 °C,
and the pH was adjusted to between 2.8 and 3.2 with 40%
NaOH solution at 3-6 °C, crystallizing the product. The
resulting mixture was stirred at 5-8 °C for 2.0 h, and the
product isolated by filtration. The solids were washed with
chilled water (∼5 °C, 15 mL) and dried under vacuum below
65 °C to yield 3.72 g of 2 (PMPA•H2O, 59.1%). The solids
had an HPLC purity of 98.4%, assaying (anhydrous basis) at
91.0% w/w, with a moisture content of 5.2% w/w.
A reaction run at the same scale, without use of tetrabutyl-
ammonium bromide, provided 1 in 56.0% yield and 98.0%
purity.
Acknowledgment
Preparation of 1 and its Fumarate Salt (TD and TDF).
Tenofovir (2, 53.1 g, 0.174 mol as hydrate) was charged to a
reaction flask and dried at about 100 °C under vacuum (25-40
Torr) for 36 h (theoretically providing 50.0 g anhydrous 2). To
this was added dry NMP (200 mL) and cyclohexane (150 mL).
The latter was removed by distillation under reduced pressure
at not more than 60 °C. Cyclohexane (150 mL) was again added
and the distillation repeated. The concentrate was cooled to 45
We thank the Bill and Melinda Gates Foundation, DFID,
and other Clinton Foundation donors for financial support.
Supporting Information Available
HPLC methods with chromatograms. This material is
available free of charge via the Internet at http://pubs.acs.org.
Received for review May 13, 2010.
OP1001337
(11) Given the large volumes of ethyl acetate required for successful
workup, recycling of the recovered solvent is indicated.
Vol. 14, No. 5, 2010 / Organic Process Research & Development
•
1201