Originally developed by AstraZeneca, gefitinib (Iressa) is a small-molecule tyrosine
kinase inhibitor of the epidermal growth factor receptor (EGFR).[1] The drug was approved in 2015 by the US Food and Drug Administration (FDA) as a
first-line treatment for metastatic non-small-cell lung cancer (NSCLC) with EGFR mutations.
Worldwide, lung cancer is the most prevalent fatal cancer for both men and women.[2] In NSCLC, mutation of the EGFR tyrosine kinase domain destabilizes the kinase conformation
and affects downstream signaling pathways.[3] These disruptions stimulate cancer cell proliferation and inhibit apoptosis. Gefitinib
reversibly binds to the ATP site of the EGFR kinase domain to inhibit autophosphorylation
and signal transduction.[4]
Several syntheses of gefitinib have been described in the literature.[5]
[6]
[7]
[8]
[9] AstraZeneca’s original synthesis began with the demethylation of 6,7-dimethyoxyquinazoline-4-one
with l-methionine and methanesulfonic acid, followed by acetylation, halogenation, aniline
nucleophilic aromatic substitution (SNAr), deacetylation, and O-alkylation (Scheme [1]).[10] This six-step synthesis (10% overall yield) required chromatographic purifications
and used hazardous reagents, such as thionyl chloride, which reacts violently with
water to produce toxic fumes of sulfur dioxide and also contaminates the air very
quickly upon evaporation at 20 °C.[11]
Scheme 1 AstraZeneca synthesis of gefitinib[10]
In 2007, Reddy and co-workers reported a synthesis of gefitinib from isovanillin (Scheme
[2]).[12] The nitro group in the isovanillin-derived intermediate was reduced with sodium
dithionite, followed by treatment with N,N-dimethylformamide dimethylacetal (DMF-DMA; 1,1-dimethoxy-N,N-dimethylmethanamine) and amination with 3-chloro-4-fluoroaniline to yield the active
pharmaceutical ingredient (API). No chromatography was required, but high reaction
temperatures were needed, and DMF was used in large quantities in the seven-step synthesis.
Scheme 2 Reddy and co-workers’ synthesis of gefitinib[12]
More recently, Suh and co-workers reported a variant of the AstraZeneca synthesis
that used a transient-protective-group strategy (Scheme [3]).[13] An acetylated quinazoline core was subjected to a chlorination with POCl3, substitution with 3-chloro-4-fluoroaniline, and deprotection with LiOH to set the
stage for alkylation with 4-(3-chloropropyl)morpholine, using TMSI to protect the
aniline nitrogen transiently. Although high yielding, this synthetic route required
hazardous TMSI and a more-elaborate starting material. It also used phosphoryl chloride,
which reacts violently with water to produce toxic gases, and is highly corrosive.[14]
Scheme 3 Suh and co-workers’ synthesis of gefitinib[13]
We envisioned a new route to gefitinib with fewer than five steps from inexpensive
starting materials that would avoid hazardous reagents and chromatographic separations,
and would keep reaction temperatures in the 0–60 °C range. Such a process would be
commercially relevant and potentially attractive for pharmaceutical manufacturing.
To increase the electrophilic reactivity of the pyrimidine moiety in the SNAr reaction, we chose commercially available 2,4-dichloro-6,7-dimethoxyquinazoline
(1) as a starting material. To the best of our knowledge, a synthesis of gefitinib or
related analogues that utilizes a 2,4-dichloroquinazoline as a starting material or
advanced intermediate is unprecedented. We reasoned that the SNAr substitution of the chlorine in the 4-position of the quinazoline would occur preferentially,[15] and that the 2-position might be readily dechlorinated at a late stage. We did not
employ a Buchwald–Hartwig amination of the quinazoline because of concerns regarding
the harsh conditions often required and because of the risk of contaminating the API
with Pd. Furthermore, several groups have recently demonstrated the feasibility of
nucleophilic aromatic substitutions on similar quinazoline substrates under simple
acidic conditions.[16]
[17]
[18]
Accordingly, the dichloroquinazoline 1 was treated with 3-chloro-4-fluoroaniline in 20.4 equivalents of acetic acid at 55 °C
for two hours to yield the coupling product 2 after extraction with EtOAc and filtration (Scheme [4]). Under these conditions, we were able to isolate the desired 4-aminated product
2 exclusively in 65% yield on a multigram scale. Not unexpectedly, however, the ensuing
selective demethylation of 2 proved challenging (Table [1]). A variety of conditions were tested, including l-methionine in methanesulfonic acid (Table [1], entry 1). However, these conditions mainly afforded decomposition products at the
high temperatures that proved necessary for significant conversion. Interestingly,
BBr3 provided the bisdemethylated product exclusively (entry 2). When we experimented
with various additives to BBr3 to control the rate of demethylation, we either observed no reaction (NR) or a complex
mixture of products (entries 3 and 4). Another Lewis acid, aluminum iodide (AlI3), also showed no reaction at low temperatures (entry 5). Aluminum chloride (AlCl3) showed a robust rate of conversion but, even in the presence of sodium iodide (NaI),
at best provided a 1:1 ratio of demethylated isomers that were difficult to separate
(entries 6–7). With ethanethiol as an additive, a favorable 1:0.4 ratio was obtained,
but in low yield (entry 8). Reaction times longer than two days were required for
high conversions.
Scheme 4 Dichloroquinazoline as a new starting material for gefitinib preparation
Table 1 Screening of Reagents and Conditions for Demethylation at the 6-Position of Intermediate
2
|
|
Entry
|
Reagent(s) (equiv)
|
Temp (°C)
|
Solvent
|
Product(s)a
|
|
1
|
l-methionine (1.2)
|
150
|
MsOH
|
– (dec.)
|
|
2
|
BBr3 (3.0)
|
r.t.
|
CH2Cl2
|
7
b
|
|
3
|
ZrCl2 (2.0), BBr3 (1.0)
|
50
|
CH2Cl2
|
NR
|
|
4
|
TiCl4 (2.0), BBr3 (1.0)
|
45
|
CH2Cl2
|
mixture
|
|
5
|
AlI3 (1.5), PhSH (1.5)
|
0
|
CH2Cl2
|
NR
|
|
6
|
AlCl3 (3.0)
|
r.t.
|
CH2Cl2
|
3, 6
c
|
|
7
|
AlCl3 (3.0), NaI (3.0)
|
r.t.
|
CH2Cl2
|
3, 6
|
|
8
|
AlCl3 (3.0), EtSH (2.0)
|
40
|
CH2Cl2
|
3, 6
d
|
|
9
|
[TMAH][Al2Cl7] (3.0)
|
50
|
CH2Cl2
|
3, 6
e
|
a Products and product ratios were determined by LC/MS and 19F NMR analyses.
b The bisdemethylated product 7 was formed exclusively.
c A 1:1 ratio of demethylated isomers and the starting material was detected that could
not readily be enriched in the desired product through crystallization.
d A 1:0.4 ratio of phenols 3 and 6 was formed in low yield.
e A 1.3:1 ratio of phenols 3 and 6 was enriched to a 97:3 ratio in the first crystallization batch favoring the desired
product 3.
In an attempt to accelerate the reaction without recourse to excessive heating that
could potentially lead to deamination byproducts and quinazoline ring opening, we
explored the cleavage of methyl ethers through the use of ionic liquid (IL) reagents,
including trimethylammonium heptachlorodialuminate {[TMAH][Al2Cl7]}.[19] The IL demethylation mechanism is similar to that of AlCl3; however, the IL contains a higher concentration of chloride ions, the nucleophilicity
of which is enhanced, resulting in shorter reaction times. Furthermore, improved demethylation
selectivity has been reported for bicyclic ring systems.[19] The IL was synthesized in situ from aluminum trichloride and trimethylammonium chloride
in dichloromethane, and was directly used for the demethylation step in a one-pot
protocol. With intermediate 2, we found that treatment with [TMAH][Al2Cl7] at 50 °C for two hours gave a 1.1–1.3 to 1 ratio of monodemethylated regioisomers;
however, a favorable >95:5 ratio of the desired product could readily be obtained
in 30–35% yield without chromatography by crystallization of the concentrated reaction
mixture from hot methanol. Although not required for the next step, a second crystallization
increased the regioisomeric purity to >99%.
The IL was freshly prepared before each use, and was not concentrated as suggested
in the original publication,[19] because we found that removal of the solvent generally resulted in a less active
reagent. The one-pot protocol also simplified the experimental protocol. Significantly,
the synthesis of [TMAH][Al2Cl7] IL is cost effective, and its feasibility for chemical-process applications has
already been demonstrated on 7 kg scale.[20]
Previous syntheses mainly used DMF, sodium and potassium carbonates, and high temperatures
for the O-alkylation step. We found that sodium and potassium carbonates were not
effective at low temperatures in DMSO. In contrast, the reaction of 3 with 4-(3-chloropropyl)morpholine in the presence of cesium carbonate in DMSO at
40 °C for 2.5 hours provided ether 4 in 80% yield after filtration and crystallization from hot methanol. The FDA classifies
DMF as a more-hazardous Class 2 solvent, whereas DMSO is a less-hazardous Class 3
solvent; therefore, these conditions were in agreement with our goal of minimizing
the use of toxic or controlled reagents.[21]
The final dehalogenation step in the conversion of 4 to 5 required considerable optimization. Palladium(II) acetate in the presence of hydrogen
gas provided no control of selectivity and resulted in complete dechlorination to
8 (Figure [1]), as well as loss of aniline, among other side reactions. Similarly, hydrogenation
with 10% Pd on carbon resulted in the loss of the quinazoline chlorine as well as
the aniline chlorine atoms to give predominantly 8. Attempts at hydrogenation by using Lindlar’s catalyst or Raney nickel both provided
only trace conversions to the desired product 5, along with varying amounts of the bisdechlorinated derivative 8. Similarly unsatisfactory results were obtained by using a combination of NaBH4 and TMEDA under palladium catalysis.[22] However, a chemoselective conversion was finally realized by using zinc and acetic
acid in the presence of tetrabutylammonium bromide (Bu4NBr) as an additive, a reagent combination that had previously been shown to reduce
a chloropyridine substrate selectively.[23] In addition to the desired product 5, under these conditions we also observed that small amounts of 3-chloro-4-fluoroaniline
were formed, which probably originated from an acid-mediated solvolysis of the quinazoline
moiety. This side reaction could be suppressed through a modification employing zinc
and N,N,N′,N′-tetramethylethylenediamine (TMEDA) in a mixture of MeOH and AcOH to achieve the
desired dehalogenation of 4 exclusively. After stirring the reaction mixture at 40 °C for 24 hours, 2-mercaptonicotinic
acid was added to assist in the removal of excess zinc and zinc salts.[24] Finally, we were able to crystallize the product from hot MeOH to provide gefitinib
(5) as colorless crystals in 82% yield with >99% purity as determined by LC/MS analysis.[25]
Figure 1 Products observed in the dechlorination of 4
In conclusion, a gram-scale synthesis of gefitinib was accomplished in four steps
from commercially available 2,4-dichloro-6,7-dimethoxyquinazoline. Reaction temperatures
did not exceed 55 °C, and workup procedures took advantage of the superior crystallization
properties of the C(2)-chlorinated quinazolines in methanol. Thus, all purifications
were performed by filtrations or crystallizations. No protective groups were required,
and reagents such as DMF, SOCl2, POCl3, and TMSI were avoided. A new application of an ionic liquid streamlined the demethylation
step, and a selective dehalogenation by using zinc, acetic acid, and TMEDA proved
successful.
