PPARγ-sparing thiazolidinediones as insulin sensitizers. Design, synthesis and selection of compounds for clinical development

Full text of DOI:10.1016/j.bmc.2018.10.033

Accepted Manuscript
PPARγ-Sparing Thiazolidinediones as Insulin Sensitizers. Design, Synthesis
and Selection of Compounds for Clinical Development
Steven P. Tanis, Jerry R. Colca, Timothy T. Parker, Gerald D.Artman III, Scott
D. Larsen, William G. McDonald, Robert C. Gadwood, Rolf F. Kletzien, James
B. Zeller, Pil H. Lee, Wade J. Adams
PII:
S0968-0896(18)31564-5
https://doi.org/10.1016/j.bmc.2018.10.033
BMC 14592
DOI:
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
5 September 2018
20 October 2018
27 October 2018
Please cite this article as: Tanis, S.P., Colca, J.R., Parker, T.T., D.Artman, G. III, Larsen, S.D., McDonald, W.G.,
Gadwood, R.C., Kletzien, R.F., Zeller, J.B., Lee, P.H., Adams, W.J., PPARγ-Sparing Thiazolidinediones as Insulin
Sensitizers. Design, Synthesis and Selection of Compounds for Clinical Development, Bioorganic & Medicinal
Chemistry (2018), doi: https://doi.org/10.1016/j.bmc.2018.10.033
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

PPAR-Sparing Thiazolidinediones as Insulin Sensitizers. Design, Synthesis and Selection of Compounds
for Clinical Development
Steven P. Tanis; *a Jerry R. Colca;b Timothy T. Parker;c Gerald D.Artman III;c,d Scott D. Larsen;e William G.
McDonald;b Robert C. Gadwood;c Rolf F. Kletzien;b James B. Zeller;f Pil H. Lee,e Wade J. Adamsg
a) SPTanis PharmaChem Consulting LLC, Carlsbad, CA 92011; b) Metabolic Solutions Development Co.,
Kalamazoo, MI 49007; c) Kalexsyn, Kalamazoo, MI 49008; d) Current address: Celgene, Summit, NJ 07901; e)
Vahlteich Medicinal Chemistry Core, University of Michigan, Ann Arbor, MI 48109; f) Jim Zeller Consulting,
Scottsdale AZ 85259; g) ADaMET Consulting LLC, Kalamazoo, MI 49008
Declaration of Interest: JRC, WGM, RFK have an interest in Metabolic Solutions Development Co.
1. INTRODUCTION
Insulin sensitizers, particularly pioglitazone, are among the most useful treatments for type 2 diabetes as they
correct the key feature of the disease and exert durable actions not only restoring the action of peripheral tissues
to insulin but also preserving the function of pancreatic beta cells that produce and secrete insulin.
Unfortunately, the mechanism of action of these molecules has been poorly understood and no new insulin
sensitizers have been approved for clinical use since pioglitazone in 1999.1 It has become generally accepted
that the transcription factor PPAR is the direct molecular target for these molecules even though pioglitazone,
the weaker PPAR activator of the two remaining thiazolidinediones (TZDs; pioglitazone and rosiglitazone) has
the clinically improved therapeutic activity vs rosiglitazone and PPAR directed efforts have yet to yield a new
TZD or a new class of insulin sensitizer with market approval. We have suggested that an alternate approach to
this problem might be to go 180 degrees away from this PPAR driven paradigm and to examine the impact of a
minimization of the direct interaction with PPAR while maintaining activity with the recently identified
mitochondrial target of these molecules.2 This approach has given rise to two new analogs in clinical
development3 and to the identification of the mitochondrial target as the mitochondrial pyruvate carrier. This
work suggests that the transcriptional networks are downstream of a metabolic event and thus might reconcile
the importance of the transcription networks to some of the pharmacology of the TZDs and the fact that direct
activation of these networks may not be a key feature of a successful drug candidate.4 Here we describe the

medicinal chemistry approach to design away from the direct activation of PPAR that was employed in an
attempt to improve on structures which had previously demonstrated clinical effectiveness.
A short history of the background endeavors, which we have previously reviewed (vide infra), is presented here.
The original biological analyses for all of the initial TZD analogs was done in diabetic animal models as
pioneered by Sohda, Ikeda, Fujita and colleagues at the Takeda company in the 1970s.5 These compounds were
found to lower both circulating glucose and insulin and they came to be known as insulin sensitizers.6 The
Takeda analogs, ciglitazone 1 (Figure 1) followed by pioglitazone 2 moving first into clinical trials, however
®
troglitazone 4, an analog synthesized by Sankyo, was the first compound from this class (as Rezulin )
®
approved for the market in 1997. This was followed by the approval of rosiglitazone 3 (as Avandia ) and then
®
pioglitazone 2 (as Actos ) in 1999.7 The PPAR IC50 numbers shown in Figure 1 were determined years after
the selection of these compounds for development as the identification of this transcription factor as a binding
site came some 10 years after the selection of the only three compounds with this action to be approved and
marketed to treat type 2 diabetes. The conclusion that peroxisome proliferator-activated receptor  (PPAR) was
the direct target of these molecules was primarily based on the activity of the most potent activator of this
original group of compounds, rosiglitazone 3.8 The conclusion that PPAR was the relevant target was bolstered
by the ability to find other compounds that activated PPAR and which were also active as insulin sensitizers.9

Figure 1: Initial thiazolidinedione insulin sensitizers- discovered though whole animal assays
Once troglitazone 4 was removed from the market because of a unique idiosyncratic hepatotoxicity, a concerted
effort was made to expand out of this chemical class using rosiglitazone 3 and related analogs as templates in an
effort to find new PPAR agonists. These efforts resulted in the discovery of many other TZD and related
molecules that have entered into development, although none of these compounds has achieved market
approval.10,11,12 The primary intent of these later chemical programs was to increase potency, specifically,
designing toward more potent and/or more selective direct activation of PPAR.13 Again, it must be noted that
the marketed clinical candidates were selected 10 years before the PPAR hypothesis was developed.
Figure 2 shows that the starting point for many of the non-TZD clinical candidates that came from industry
programs was a Takeda compound AD-5075, one of the most potent analogs from the original medicinal
chemistry approach in terms of ability to stimulate the differentiation of adipocytes.14 Potent PPAR agonists
were created by replacing the TZD ring with other acidic bioisosteres (the calculated pKa of the majority of the
isosteres ranged between 6.4 for TZDs to 3.6 for alkoxy acids), with most of the efforts keeping the left-hand
portion of the molecule fairly constant, while deleting the OH-from the linker.15-20 Some of these changes
resulted in analogs that bear resemblance to the fibrates (which are known as PPAR agonists) that also
activated other nuclear receptors from the PPAR family. These have included compounds that affect some (, ,
or +) or all (, , ) of the PPAR-related nuclear receptors.21 The one constant for both the newer TZDs and
the non-TZDs that have failed (summarized on the right side of Figure 2) is that they were all designed with the
ides of activating PPAR and then particularly with the non-TZDs, other members of the PPAR class.

Figure 2: Insulin sensitizer families. Ciglitazone was the first insulin sensitizer tested in clinical trials.
Troglitazone, rosiglitazone, and pioglitazone are the only insulin sensitizers ever to be approved for treatment of
type 2 diabetes. Representative non-TZDs are shown on the right side of the page. Most of these compounds
were modeled after Takeda’s AD-5075, which was the most potent of the original Takeda analogs at
differentiating 3T3L1 preadipocytes,14each with various replacements for the TZD ring. None of these
compounds have been approved in the major markets.
The case for a prominent role of PPAR as a mainstay of the pharmacology was made mostly from study of the
differentiation of adipose tissue. However, the pleitropic insulin sensitizing pharmacology of the
thiazolidinediones cannot be solely explained by action on preadipocytes and the expansion of white fat22a,b,cas
TZD’s still exert some actions in animals wherein white fat has been eliminated.22d There are effects of these
compounds on multiple cell types that cannot be solely attributed to the direct stimulation of PPAR, a
transcription factor which drives the synthesis and storage of lipids primarily in adipose cells. There is also little

doubt that the weight gain and fluid retention, the main side effects that have limited the use of pioglitazone and
rosiglitazone, are clearly due to PPAR agonism.
As we have reviewed elsewhere, the limited success of these abovementioned endeavors might be due to the
focus of all research activities on the wrong target(s).22c There is evidence that a considerable portion of the
insulin-sensitizing effect might involve anti-inflammatory or direct metabolic actions, including those in the
mitochondrion.22b It should also be pointed out that recent evidence from long-term trials has suggested that of
the two TZDs currently in clinical use, the weaker of the PPAR activators, pioglitazone 2,2,23 has significant
advantages over the stronger activator, rosiglitazone 3, in practice (Figure 1).24-26 This includes recent evidence
that long-term treatment with pioglitazone might prevent both dementia27a as well as favorably affect
cardiovascular outcomes including heart attacks and stroke.27b,c These findings together with the failure of
multiple efforts to successfully develop structurally diverse, potent PPAR activators for the treatment of
diabetes support the view that a diminution of direct chemical activation of this transcription factor should be
pursued.22c This hypothesis can be more directly tested with a series of TZDs that avoids binding to PPAR
transcription factors. If this hypothesis is correct, insulin-sensitizing pharmacology as well as other pleiotropic
effects of TZDs could result from defined, direct action on non-PPAR targets and this should allow the insulin-
sensitizing pharmacology to be unencumbered by PPAR-related side effects.
As we have described elsewhere, evidence has been accumulating suggesting, that TZDs have direct effects on
the mitochondria from various types of cells.22a, 28 Labeling studies with tritiated pioglitazone29a and a
photoaffinity pioglitazone-analog29b-cled to proteins in the inner mitochondrial membrane which were initially
called the mitochondrial target of thiazolidinediones (mTOT). Further studies resulted in the identification of
these proteins as the mitochondrial pyruvate carrier (MPC), composed of two proteins MPC1 and MPC2 in the
inner mitochondrial membrane.30a-c This complex is required for pyruvate to enter the mitochondrial matrix,
where it is metabolized, from its site of synthesis in the cytosol. Recent efforts have suggested targeting MPC
for treatment of a variety of metabolic and inflammatory diseases, including diabetes,23, 31 Parkinson’s and other

neurodegenerative diseases,32and nonalcoholic fatty liver disease (NAFLD).30a, 33
Recent reports have suggested that the alteration of carbon flow through the pyruvate carrier results in changes
in carbon flow from other sources31b, 34a,bin order to construct requisite building blocks, which has implications
with respect to the delivery of substances to key cellular regulatory machinery.34c These entities are the
executors of post-translational regulation of proteins and the subsequent epigenetic regulation of gene
expression. TZD treatment and the impact that these compounds have on metabolism thus results in the
differentiation of stem and progenitor cells.35 Another important aspect to consider in the discussion of the
pharmacology of the TZDs in vivo is that significant efficacy only occurs in metabolically challenged animal
models which are insulin resistant, either genetically, or by co-treatment with drugs or hormones (e.g.
glucocorticoids, growth hormone) or with diets/agents which cause insulin resistance.7, 22b Therefore, it appears
that the effect of TZDs is to reduce insulin resistance rather than to increase insulin sensitivity per se, making it
difficult to study the effects of TZDs on isolated cells in culture.
As it is generally accepted that excess calories can help predispose to insulin resistance and type-2 diabetes, and
weight loss and reduction of caloric intake can protect against even moderate diabetes, it is likely that the
precipitation of the diabetic state involves a plethora of systems that are regulated by the availability of
nutrients. Nutrient sensing mechanisms such as AMPK and the mTOR complexes coordinate downstream and
parallel control points of gene expression, and TZDs are known to impact these processes. In many systems,
treatment with TZDs has been shown to result in both activation of AMPK and reduction of the activity of
mTORC1.35 The precise mechanistic details of this effect remain to be elucidated, however one could imagine
that selective signals could be generated as a function of the relative amount of pyruvate that enters through the
facilitated carrier vs. the alternate delivery of substances, and that the understanding of these downstream events
will afford important insights into treating metabolic disease. With this understanding, in 180 degree opposition
to the prevailing hypothesis, we sought to reduce the direct interaction of new analogs with PPAR.

The initial design criteria that informed this effort to reduce the direct activation of PPAR was provided by a
simple two-dimensional overlay of the key thiazolidinediones 1 – 4 (Figure 3). The shortest backbone is
associated with ciglitazone 1 (red) which also is the least effective PPAR activator (IC50 >5M). The next
longest backbone extension is exhibited by pioglitazone 2 (PPAR IC50 = 1.79M) with rosiglitazone 3 (PPAR
IC50 = 0.112M) and troglitazone 4 (PPAR IC50 = 0.370M) pushing relatively lipophilic aryl rings further
from the thiazolidinedione head group than 2. This extension of planar lipophilicity appeared to be of some
significance with respect to PPAR activity, but it seemed to be abrogated by the presence of polarity in the
vicinity of the lipophilic terminus.
Figure 3: 2D Overlays of Glitazones 1 - 4
If we compare the 2D overlay of pioglitazone 2 (PPAR IC50 = 1.79M) with muraglitazar 536 (PPAR IC50 =
120nM) we see (Figure 4) an extension of lipophilicity beyond that associated with rosiglitazone 3 or

troglitzaone 4 with an imperfect overlay of the acid moiety with the TZD. In the case of muraglitazar 5 a
PPAR IC50 = 120nM does support the notion of the impact of backbone length and lipophilicity of PPAR IC50
values.
Figure 4: 2D Overlay of a TZD – Pioglitazone 2 with a non-TZD Muraglitazar.
As a result of the general SAR discussed above (vide supra) we elected to consider the preparation of
thiazolidinediones of the same relative backbone length as that of pioglitazone with polarity near the terminus in
the form of the pyridine-N and a proximal C=O or CH-OH as well as a benzene ring, alkyl- or alkoxy-
substituted and with a proximal C=O or CH-OH. These entities would be limited in number and in relatively
close physicochemical space to an entity, pioglitazone 2, of clinical interest / importance. These notions would
allow us to further examine the impact of polarity near the terminal aromatic ring as well as the potential for
asymmetry, in the form of the CHOH, to alter the PPAR IC50. This led us to design the series of 12 molecules
(5-16) shown in Figure 5 below. The ketone targets are presented in descending order of their c-LogD values.37

Figure 5: Targets 5 – 16 Designed to Examine the Impact of Polarity on PPAR and Antidiabetic Activity
2. CHEMISTRY
We had previously reported the synthesis of 638starting from Pioglitazone 2. Of note in that endeavor, was the
difficult oxidation of the carbinol precursor of 6.38,39 Ketone 8 was then targeted using a styrene oxide approach
(Scheme 1)40 developed during a process chemistry evaluation of the synthesis of 6.41 This initial epoxide ring
opening reaction encountered similar selectivity issues as had been described40 and a somewhat lower yield was
realized. Exposing styrene oxide to the phenoxide prepared from 4-hydroxybenzaldehyde in a toluene /
PEG4000 mixture led to a 1.8:1 mixture of 20:19 with 20 isolated in 34% yield. In the case of the pyridyl based
epoxide40 a ca. 4:1 ratio of primary vs secondary attack was realized with the desired material comprising ca.
80% of the mixture. Clearly, this styrene oxide system performed differently, suggesting the need for alternate
approaches to 5 and 7.

Scheme 1. (a) PEG4000, 1M aq. NaOH, toluene 80C, 34% [20]; (b) 2,4-thiazolidinedione, piperidine, abs,
EtOH , 86%; (c) CoCl2-6-H2O, dimethylglyoxime, 1M aq. NaOH, aq. NaBH4 (in 0.2M aq. NaOH), 76%; (d)
DMSO, P2O5, Et3N, CH2Cl2 (40%).
With 20 in hand we continued our efforts toward 8 (Scheme 1) as we have previously reported.38 Condensation
between 20 and 2,4-thiazolidinedione, mediated by piperidine, gave the expected Knoevenagel product (86%),
which was smoothly reduced using the modified conjugate reduction protocol (CoCl2-6H2O, dimethylglyoxime,
NaBH4)38 to give racemic alcohol 21 (76%). Application of the Taber39 oxidation conditions38 to 21 did lead to
the desired 8, albeit in a disappointing 40% yield.
Our next target, the 4-ethyl-substituted ketone 5, (Scheme 2) would serve as the test case for direct ketone
introduction and an examination of overall yield for the redox cycle at that center required as a result of the
chemistry utilized for the conjugate reduction of the methylene-2,4-thiazolidinedione bond. In the event the
known bromide (R1 = CH2CH3, R2 = H was readily reacted with 4-hydroxybenzaldehyde (K2CO3, acetone) to
afford aldehyde 22a (61%). Knoevenagel condensation (2,4-thiazolidinedione, piperidine) afforded 23a (55%)
which, when exposed to conjugate reduction accomplished with 2,2’-bipyridine as the cobalt ligand, led to the
expected doubly reduced alcohol 24a (76%). Application of the Taber oxidation conditions to this ethyl
substituted benzylic alcohol resulted in and a 72% yield of ketone 5. The overall yield of 5 as shown in Scheme
2 was a modest 18%.

Scheme 2. (a) 4-hydroxybenzaldehyde, K2CO3, acetone, 22a 61%, 22b 62%; (b) 2,4-thiazolidinedione, abs.
EtOH, piperidine, , 23a 55%, 23b 76%; (c) CoCl2-6H2O, 2,2’-bipyridine, 1N aq. NaOH, NaBH4 24a 76%, 24b
85%; (d) from 24a P2O5, DMSO, i-Pr2NEt, CH2Cl2, 5 72%; from 24b IBX, EtOAc, , 7 77%.
The final ketone target 7 was prepared as shown in Scheme 2. Commercially available 3-methoxy-phenacyl
bromide was coupled with 4-hydroxybenzaldehyde to give 22b (62%) and Knoevenagel condensation then
furnished 23b (76%). Conjugate reduction and concurrent ketone reduction then led to alcohol 24b in a
reasonable 85% yield. Benzylic alcohol oxidation was accomplished with IBX (refluxing EtOAc) to yield
ketone 7 (77%). IBX had been examined during the extensive oxidation studies of Scheme 138 and again during
the process evaluation41 to no avail in that pyridyl system. The direct ketone introduction, redox paradigm of the
oxidation modification noted in Scheme 3 did provide 7 in a somewhat improved 31% yield when compared
with the Scheme 3 results leading to 5 wherein the Taber oxidative paradigm was employed.
3. RESULTS AND DISCUSSION
With our target ketones 5-8 in hand we planned to allow their evaluation in vitro and in vivo inform us with
respect to those ketones which might then be selected for chiral reduction to alcohols such as those depicted in
Figure 5. Ketones 5-8 were first examined in a PPAR TR-FRET competitive binding assay42a using
rosiglitazone 3 as the positive control which was employed to measure binding affinity for PPAR. All of the

ketones of Table 1 show diminished PPAR binding affinity when compared to pioglitazone 2. Pioglitazone
derived 638 (PPAR IC50 = 22.30M) and methoxy-aryl ketone 7 (PPAR IC50 = 18.14M), shifted ca. 13X and
10X from 2 respectively, exhibited the poorest PPAR binding affinity. Thus, ketones 6 and 7 were selected for
in vivo evaluation38, 43 of antidiabetic activity in the obese, hyperglycemic, hyperinsulinemic, KKAy mouse.43d
Animals were grouped into treatment and control groups through pre-test blood glucose measurements and
treatment groups had ketones 6 and 7 administered as a food admixture at 100mg / kg / day for 4 days.43d The
glucose level for the treated group (T) over the control group (C) was used to determine the antihyperglycemic
activity and the results are reported as a T/C value. This is computed by the ratio of the terminal glucose level of
mice that have undergone 4 days of treatment with the experimental compound divided by the average glucose
level of the control group given the vehicle control for the study. A compound with a T/C value ≤ 0.85 is
considered active in this model of diabetes. Pioglitazone 2 (100mg / kg) served as the positive control.
Pioglitazone 2, the positive control lowered blood glucose in the KKAy mice providing a T/C = 0.56. Ketones 6
and 7, which exhibited 12.5X and 10X poorer PPAR binding selectivity than 2 were fully effective as
antihyperglycemic agents and were associated with T/C values of 0.52 and 0.58 respectively.
Table1: PPAR Binding Activity (IC50) and Antihyperglycemic Activity of Ketones 5-8
PPAR IC50
(M)
KKAy
T/C
Compound
1.79
0.56
NT
8.46
22.30
0.52

18.14
11.70
0.58
NT
The data of Table 1 suggested that ketones 6 and 7 would serves as appropriate substrates for conversion into
the respective enantiomeric alcohols 11, 12 and 13, 14 (Figure 5) for evaluation of this change in polarity and
the presence of asymmetry near the terminus on PPAR binding and antihyperglycemic activity. Toward that
end 6 and 7 were reduced using our previously reported modified Noyori conditions44 as shown in Scheme 4 to
afford alcohols 11, 12, 13, and 14 in good to excellent yields (85-98%) and excellent enantiomeric excess at the
carbinol center (96%ee). We were not concerned about the asymmetric center at the 5-position of the
thiazolidinedione moiety as it has been demonstrated to readily racemize in vivo.43c
Scheme 3: (a) i. dichloro(6-p-cymene)ruthenium(II) dimer, (1R, 2R)-(-)N-p-tosyl-
1,2,diphenylethylenediamine, iPrOH, ; ii. 6, HCOOH/Et3N, 85%, 96%ee; 7 98%, 96%ee; (b) i. dichloro(6-p-
cymene)ruthenium(II) dimer, (1S, 2S)-(+)N-p-tosyl-1,2,diphenylethylenediamine, iPrOH, ; ii. 6,
HCOOH/Et3N, 90%, 96%ee; 7 96%, 96%ee.

Table 2 presents the data associated with rosiglitazone 3, pioglitazone 2, ketones 6, and 7 as well as the related
enantiomeric alcohols 11, 12, 13, and 14 gathered in the aforementioned in vitro PPAR binding assay (IC50)42a,
the PPAR direct transactivation-agonist assay (EC50)42b, as well as the antihyperglycemic activity (T/C) as
measured in the KKAy mouse model.38,43 Starting from the benchmark compounds rosiglitazone 3 and
pioglitazone 2, the aforementioned binding affinities (3 IC50 = 0.112M; 2 IC50 = 1.79M) illustrate that
rosiglitazone 3 is a ca. 16X better PPAR binder than pioglitazone 2. Although the absolute values for EC50 -
IC50 differ in the transactivation assay as compared to the binding assay, the relative differential potencies of the
compounds mirrored fairly well in the PPAR direct transactivation-agonist assay with 3 exhibiting an EC50 =
0.008M while pioglitazone is associated with EC50 = 0.098M, a >12X differential. Both 3 and 2 are fully
effective as in vivo antidiabetic agents with KKAy T/C values of 0.55 and 0.56 respectively. Ketones 6 and 7
are poorer PPAR binders than 3 and 2 (6 200X vs 3; >12X vs 2; 7 162X vs 3, 10X vs 2). The PPAR
transactivation-agonist activity EC50 values measured for 6 (EC50 = 0.657M) and 7 (EC50 = 0.961M) are
shifted ca. 7-10X vs 2 and 80-100X vs 3. As mentioned previously (Table 1), ketones 6 and 7 exhibit good
antihyperglycemic activity (T/C = 0.52, 0.58 respectively), equivalent to that of 3 and 2. The impact of chiral
reduction on the PPAR binding activity (IC50) and PPAR transactivation-agonist activity (EC50) of the
alcohols thus formed is both interesting and surprising. The physicochemical space occupied by ketone vs
alcohol as determined by the cLogD value changes little for the pyridyl system (6 cLogD = 1.75; 11/12 cLogD
= 1.87), and alters to a greater extent for the phenyl congeners (7 cLogD = 1.60; 13/14 cLogD = 1.16). The
magnitude of these changes might be expected to have a minor effect on the PPAR binding activity, however
as is obvious from the data of Table 2, asymmetry at the carbinol center has a very large impact on PPAR
binding IC50 and PPAR transactivation-agonist EC50 values. The reduction of the carbonyl of ketone 6 to S-
alcohol 11 results in a >7x shift in binding activity (6 PPAR IC50 = 22.30M to 11 PPAR IC50 = 3.12M)
toward stronger binding and a 2.4X shift toward greater activity in the PPAR transactivation-agonist assay (6
EC50 = 0.657M to 11 EC50 = 0.271M) while reduction to the R-alcohol 12 has the opposite impact shifting
toward weaker PPAR binding affinity (12 PPAR IC50 = 50.14M) a shift of more than 16-fold 12 vs 11, and a

2.6X shift toward poorer PPAR transactivation-agonist activity (6 EC50 = 0.657M, 12 EC50 = 1.74M).
Similar carbinol impacts are observed for the S-alcohol 13 (PPAR IC50 = 2.32M; PPAR EC50 = 0.172M)
derived from 7 (PPAR IC50 = 18.14M; PPAR EC50 = 0.961M) while the R-alcohol 14 is shifted some 34-
fold toward weaker binding / ca. 2X shift toward poorer transactivation-agonist activity, exhibiting very little
binding affinity (14 PPAR IC50 = 78.45M) and lesser transactivation-agonist activity (PPAR EC50 =
1.72M). As this progression from alkane to ketone to R-alcohol has resulted in successively poorer PPAR
binding / lower PPAR transactivation-agonist activity we would expect a corresponding falloff in
antihyperglycemic activity, as measured by the KKAy mouse T/C if PPAR binding and agonist activity were
critical for blood glucose lowering in vivo. Ketones 6, 7 and alcohol 12 demonstrated no significant binding to
PPAR or PPAR using GW7647 and GW0742 as positive controls respectively. Full antihyperglycemic
activity (Table 2) was still observed for 6, 7, 12, and 14. Clearly, designing away from potent, direct activation
of PPAR has not been detrimental with respect to antihyperglycemic activity. Perhaps these data instead
implicate the importance of an alternate target, supporting the possible direct effects on the aforementioned
mitochondrial target mTOT (vide supra) with an impact on pyruvate transport. All of the active compounds
were found to compete with the binding of the photoprobe which identified MPC2.29b The IC50’s for compounds
6 and 7 to displace the photoprobe from MPC2 are approximately 1.2M. This is similar to the apparent
affinities of the approved diabetes drugs pioglitazone 2 and rosiglitazone 3 in the same assays. Furthermore, all
of the active TZD’s also slowed the entry and mitochondrial oxidation of pyruvate.45
Table 2: PPAR Binding Activity (IC50), PPAR Direct Transactivation-Agonist Activity (EC50), and
Antihyperglycemic Activity of Pioglitazone 2, Ketones 7-8, and Alcohols 11-14
PPAR
IC50 (M)
PPAR
EC50 (M)
KKAy
T/C
Compound
0.112
0.008
0.55

1.79
22.30
3.12
0.098
0.657
0.271
0.56
0.52
0.54
50.14
18.14
2.32
1.74
0.961
0.172
0.56
0.58
0.53
78.45
1.72
0.65
To rationalize the observed SAR for PPARγ binding affinity of our new analogs several X-ray crystal structures
of PPARγ complexed with ligands were examined, and molecular docking studies were carried out using
MOE.46a There are X-ray crystal structures of the PPARγ ligand binding domain (LBD) with a number of
members of the thiazolidinedione (TZD) class of compounds such as lobeglitazone (5y2t.pdb)47, rosiglitazone 3
(4ema.pdb)48a, pioglitazone 2 (5y2o.pdb)47, and rivoglitazone (5u5l.pdb)48b. The detailed analyses of PPARγ
LBD and its interaction with ligands are described in the literature47. In brief (Figure 6), the LBD has a large Y-
shaped pocket, which consists of three sub-pockets. The middle pocket is surrounded by mostly nonpolar
residues, while the AF-2 pocket, depicted to the right in Figure 6, has a number of polar residues making
multiple hydrogen bonds with the TZD head group of ligands, and the pocket at the other end of the Y, the Ω-

pocket, has mainly hydrophobic residues with a few polar residues. When these structures are superimposed, the
binding sites near ligands are virtually identical with each ligand binding in a similar fashion. In fact, the TZD
in each compound is exactly superimposable, shown bound in the preferred S-antipode for the TZD. While the
LBD of the nuclear hormone receptors is known to be flexible, the structures of the PPAR LBD complexed
with lobeglitazone, rosiglitazone, pioglitazone, and rivoglitazone are very similar to each other. The root mean
square deviations (RMSD) for the C atoms of 260 residues of the A chains are 0.58 Å, 0.82 Å, 0.84 Å, and
0.92 Å between complexes with lobeglitazone and pioglitazone, between rivoglitazone and pioglitazone,
between lobeglitazone and pioglitazone, and between pioglitazone and rosiglitazone respectively. The structures
of the residues within 5 Å from the ligands are more similar. The residues in the middle pocket are very similar,
with the exception of F363. The phenyl ring positions of F363 vary slightly, but all remain involved in -
interactions with the ligand. The E259 and R280 residues, at the end of the Ω-pocket are most flexible, but they
are remote with respect to the ligands, rosiglitazone and pioglitazone. The compounds of interest are
structurally close to pioglitazone and do not extend to these residues. Rigid as well as flexible docking of
ligands were examined, with an induced fit option in MOE, and the docking results were compared with the X-
ray structures. Both dockings reproduced the bound conformations of the ligands in the X-ray structures, hence
rigid docking was employed. Figure 6 shows a superimposition of pioglitazone 2 (green) and rosiglitazone 3
(magenta) complexed with the PPARγ LBD. The molecular surface of the binding pocket is colored by
lipophilicity: hydrophilic in purple and lipophilic in green. Overall, pioglitazone 2 and rosiglitazone 3 bind in
very similar fashion and they superimpose very closely through the benzyl-phenyl ring and the oxo-ethyl chain.
The two ligands differ at their termini wherein the N-methyl group of rosiglitazone 3 is projected into the
lipophilic pocket of the receptor while pioglitazone 2 projects the pyridine nitrogen into this lipophilic pocket.
The remainder of the two structures maximize lipophilic interactions with the LBD. We propose that the
rosiglitazone 3 N-methyl-lipophilic pocket interaction vs the pioglitazone 2 pyridine-N projection into this
lipophilic pocket largely accounts for the lower PPAR IC50 of 0.112 µM for rosiglitazone 3 compared to the
PPAR IC50 = 1.79 µM of pioglitazone 2.

Docking studies of 12 compounds, 1-4, 5-8, and 11-14, with PPARγ IC50 data into the LBD of PPARγ were
carried out using 5y2o.pdb. As described above all of the compounds were bound in the preferred S-TZD
antipode. The top five docking poses were examined for each and the pose with the best superimposition of the
TZD with the pioglitazone 2 crystal structure was selected for comparison. The results from this 1-8
superimposition (not shown) supports the magnitude of the PPAR IC50 data presented in Figure 3 and Table 1
with respect to placement of lipophilic moieties as well as polar groups. Figure 7 presents an overlay of the
crystal structure of pioglitazone 2 and the docking pose of ketone 6. This superimposition suggests that the
polar carbonyl group of 6 is projected further into the lipophilic pocket than is the pyridine-N of pioglitazone 2,
which supports the poorer binding IC50 = 22.30 µM of compound 6 compared to the data association with
pioglitazone 2 (IC50 = 1.79 M). Ketone 7 superimposed with pioglitazone 2, models similarly (not shown) to
the 2 / 6 pairing absent the polar interaction of a pyridine-N but including the projection of the polar carbonyl
into the lipophilic pocket, also resulting in reduced PPAR binding affinity (IC50 = 18.14 M).
The significant eudismic ratio of the antipodal alcohols 11 and 12 is of intense interest and can be readily
rationalized from modeling. To that end Figure 8 shows a docking of alcohols 11 and 12 and their
superimposition with the crystal structure of pioglitazone 2. The alcohols superimpose with pioglitazone 2 very
well from the TZD-moiety through the alkoxy-ethyl chain. Compounds 2 and 12 have a good concordance of
the pyridylethyl portions of the molecule which results in the projection of the polar carbinol-OH into the side
wall of the lipophilic pocket. The antipodal S-alcohol 11, however, exhibits a slightly shifted pyridyl ethyl
moiety which allows the carbinol-OH to form a strong hydrogen bond with the carbonyl of Leu340. These
interactions allow us to rationalize the poorer binding affinity of R-alcohol 12 (IC50 = 50.14 M) as well as to
understand the stronger binding affinity associated with S-alcohol 11 (IC50 = 3.12 M). Modeling performed
with alcohols 13 and 14, superimposed with 2 (not shown) result in a similar conclusion with respect to the
poorer binding affinity of R-alcohol 14 (IC50 = 78.45 M) as well as the Leu340-H-bond driven improved
binding affinity exhibited by S-alcohol 13 (IC50 = 2.32 M).

Figure 6: Overlay of crystal structures of pioglitazone 2 (green) and rosiglitazone 3 (magenta) complexed with
the LBD of PPARγ. The molecular surface of the binding pocket is colored by lipophilicity: hydrophilic in
purple and lipophilic in green.

Figure 7: Overlay of the docking pose of ketone 6 (magenta) and the crystal structure of pioglitazone 2 (green).
Figure 8: Overlay of docking poses of S-alcohol 11 (magenta) and R-alcohol 12 (cyan) and the crystal structure
of pioglitazone 2 (green). The hydroxyl group of compound 11 is making a hydrogen bond with the carbonyl of
Leu340, while the hydroxyl group of compound 12 is turning away from Leu340 into lipophilic pocket.
The favorable characteristics associated with ketones 6 and 7 and the attendant alcohols 12 and 14 spurred us to
examine additional properties for 6 and 7 preparatory to possible selection as development candidates.
Subsequent metabolism studies (vide infra) performed on 6 and 7 would confirm the preferential formation of
desired R-alcohol metabolites 12 and 14 vs the much less desired reductive metabolism to furnish S alcohol
metabolites 11 and 13. A collection of in vitro and in vivo derived data for ketones 6 and 7 is presented in
Figure 9.
6
7
PPAR IC50 = 22.30M
PPAR EC50 = 0.657M
KKAy T/C = 0.52
PPAR IC50 = 18.14M
PPAR EC50 = 0.961M
KKAy T/C = 0.58

hERG IC50 >100M
CYP2C8 IC50 = 26M
hERG IC50 >20M
CYP2C8 IC50 = 7M
CYP2C9 IC50 = 22M
hPPB = 99%
CYP2C9 IC50 = 23M
hPPB = 99%
%F (rat) = 73%
%F (rat) = 100%
AMES (+S9) – negative
Genotoxicity – not significant
AMES (+S9) – negative
Genotoxicity – not significant
Figure 9: Collected in vitro and in vivo data for ketones 6 and 7
4. Conclusion
Compounds 6 and 7 are currently progressing through clinical development.12 Compound 6 was shown to have
similar efficacy to 45 mg pioglitazone in a 3 month phase 2b study in diabetic subjects with much less of the
PPAR-related side effects (weight gain and reduction in hematocrit) observed.31a Compound 6 was also shown
to preserve utilization of glucose in selective brain regions as determined by 18F-2-deoxyglucose PET in
subjects with Alzhiemer’s disease.32b Compound 7 is currently in development for NASH in a 52 week dose
ranging trial.30a Preclinical studies have shown that unlike direct activators of PPAR, this compound does not
have the same liabilities in terms of stimulating the reabsorption of bone.40 Also of interest with respect to
activation of PPAR are the metabolic profiles of compounds 6 and 7 found in clinical trials. In the clinic
ketone 6 (PPAR IC50 = 22.30M) was observed to be rapidly reduced to 2 metabolites, alcohols 11 (PPAR
IC50 = 3.12M) and 12 (PPAR IC50 = 50.14M) in a 10:90 ratio respectively.31a, 50 Likewise ketone 7 (PPAR
IC50 = 18.14M) was found to be rapidly metabolized in vivo to a 14:86 ratio of alcohols 13 (PPAR IC50 =
2.32M) and 14 (PPAR IC50 = 78.45M) in humans.51 As both 6 and 7 were observed to be stereoselectively
metabolized in the clinic to the R-alcohols 12 and 14, the shift toward poorer PPAR metabolite binding and
lesser PPAR residual side effects (vide supra) is noteworthy. Compounds 630a, 31b and 751 together with their
major alcohol metabolites 12 and 14 have half-lives in vivo from 10-17 hours and have been studied in clinical

trials with once daily dosing. At the clinical doses selected for 630a and 751 the AUC (area under the curve)
observed for 630a and its major metabolite 1230a, as well as the AUC measured for 751 and its major metabolite
1451, both bracket the AUC measured for pioglitazone 2 and its active metabolites at a 45mg dose.
We had previously noted (vide supra) that designing away from potent PPAR –binding / transactivation did
not result in a detrimental impact in antihyperglycemic activity in the KKAy mouse model. Of even greater
significance is the clinical data associated with ketones 6 and 7 presented above. With this collection of in vivo
data (murine and human) we have shown, for the first time, that intentionally designing compounds to have
minimal PPAR binding / transactivation activity can be accomplished without diminishing in vivo antidiabetic
activity, while greatly reducing undesired PPAR side effects. The current evidence indicates that the
pharmacology of these molecules is mediated through alteration of the function of the mitochondrial pyruvate
carrier in multiple cell types and settings.29a More recent efforts have resulted in the preparation of analogs with
PPAR IC50 >200M (binding affinity) and full antihyperglycemic activity in the KKAy mouse model (T/C =
0.55). These results will be presented in due course.
5. Experimental Section
All reagents were used as received unless otherwise stated. All reactions were performed under a blanket of
nitrogen in oven (150C) dried glassware with rigid exclusion of moisture from all reagents, solvents, and
glassware unless otherwise mentioned. Solvents for extraction and purification were of HPLC grade.
Chromatographic purification was accomplished using the flash technique of Still et al52 with the following
particulars noted: (column diameter in mm OD, quantity of silica gel employed, silica gel mesh) and eluted with
the solvents mentioned. Proton magnetic resonance spectra (1H-NMR) were recorded on a Bruker AVANCE
400 spectrometer at 400MHz in the solvent indicated. Chemical shifts are reported in parts per million ( scale)
from internal tetramethylsilane. Data are reported as follows: chemical shifts [multiplicity (s = singlet, brs =
broad singlet, d = doublet, dd = doublet of doublets, t = triplet, q = quartet, m = multiplet), coupling constant
(Hz), integration]. Most intermediates, and all final compounds were determined to exhibit >95% purity by
HPLC. HPLC analyses were performed with an Agilent 1100 HPLC equipped with a Zorbax Eclipse XDB-C18

50 x 4.6mm 1.8 micron column unless otherwise indicated. Solvent A water (0.1% TFA), solvent B acetonitrile
(0.07% TFA), gradient 5 min. 95% A to 95% B; 1 min. hold; 1 min. recycle; 30 second hold. UV detection @
210 and 250 nM with no reference. Mass spectral analyses were performed on a Waters micromass ZQ; Waters
2695 separations module.
Biological Assays
Human PPAR Competitive Binding: The assay employed a glutathione S-transerferase (GST) labelled
PPAR ligand binding domain in combination with a terbium labeled anti-GST antibody and a fluorescent
labelled small molecule PPAR tracer ligand. Competitive binding between the unlabeled competitor and the
fluorescently labelled PPAR tracer ligand for the nuclear receptor ligand binding site causes a decrease in the
FRET signal between the antibody and the tracer. Dose response curves for each test compound were prepared
by serial dilutions in DMSO. For the assay, 2X working stocks were prepared from the DMSO stocks in assay
buffer over the range from 0.001 to 250M, resulting in a final DMSO concentration of 1%. Rosiglitazone 3
was employed as the positive control, prepared as above, with a dose response curve from 0.0003 to 30M. In
384 well polypropylene plates test compound and positive control 3 treated wells received 20 L of 2X
competitor working stock, 10 L of 4X tracer ligand and 10L of 4X PPAR ligand binding domain/terbium
labelled anti-GST antibody. The negative control wells received 20L of 2% DMSO in assay buffer, 10L of
4X tracer ligand and 10mL of 4X PPAR ligand binding domain/terbium labelled anti-GST antibody. Each
concentration for all concentration groups was assayed in triplicate. The binding interaction was allowed to
reach equilibrium by incubating the plate in the dark at 20-24C for 1 hour. Fluorescence emission was
determined on a Bio Tek Synergy 2 plate reader. Two scans were obtained to determine 520/495 ratios. Scan 1
used a 360/40nm excitation filter and a 520/25nm emission filter. Scan 2 used a 360/40 excitation filter and a
495/10nm emission filter. The negative control wells were averaged for background subtraction. IC50 values
were calculated using a 4-parameter non-linear analysis using Gen5 software (V2.03.1). Dose response curves
were generated using Graph Pad Prism software (V5.0).

KKAy Mouse Assay:2,43d Groups of 8-12 week old KKAy mice were orally dosed once daily with 30 mg/kg
compound for 4 days. The compounds were given as a suspension in 0.1% Tween 80, 1% low viscosity sodium
methyl carboxycellulose. Blood glucose was measured after a four hour fast before treatment and 18-20 hours
after the last dose. The T/C ratio is the treatment (post 4 days) divided by the initial blood glucose in mg/dL.
The number shown is the average for 6 mice.
Human PPAR Transactivation – Agonist Assay: The GeneBLAzer PPAR gamma assay uses the ligand
binding domain (LBD) of human Peroxisome Proliferator-Activated Receptor-gamma (PPAR gamma) fused to
the DNA binding domain (BD) of Gal4 stable integrated into the GeneBLAzer UAS-bla HEK 293H cell line.
GeneBLAzer UAS-bla HEL 293H cells stable express a beta-lactamase reporter gene under the
transcriptional control of an upstream activator sequence (UAS). When an agonist binds to the LBD of PPAR
of the GAL4 DBD-PPAR gamma LBD fusion protein, the protein binds to the UAS, resulting in the expression
of beta-lactamase. Florescence Resonance Energy Transfer (FRET) substrate is added to the cells to generate a
radiometric dual color (blue/green) reporter signal for both stimulate and unstimulated cells. Rosiglitazone 3
was used as the positive control. Dose response curves for 3 and each test compound were prepared by serial
dilutions in DMSO covering a range of 100pm to 25M with a final DMSO concentration of 0.5%. Each
concentration was analyzed in triplicate wells. The frozen GeneBLAzer UAS-bla HEK 293H cells were
suspended in phenol red-free DMEM (Invitrogen #21063-029) containing 2% charcoal stripped FBS
(Invitrogen #12676-029) and 100 U/mL penicillin and 100g/mL streptomycin (Invitrogen #15140-122).
30,000 cells per well in 32L of medium were transferred into black, clear bottom poly-D-lysine 384 well
plates (BD, #354663) and incubated for 2 hours at 37ºC/5% CO2. Following attachment of the cells, 8L of
assay medium containing 5X drug working stocks were added. Additionally, 8L of 0.5% DMSO was added to
the unstimulated cells as well as the cell free wells. The cells were incubated for 16 hours at 37ºC/5% CO2,
treated with FRET substrate for 2 hours at room temperature in the dark and read on a Biotek Synergy 2 plate
reader. Two scans were obtained to determine 460/528 ratios. The first scan used a 400/30 nm excitation filter

and a 528/20 nm emission filter for measuring the green channel. The cell free control wells were averaged for
background subtraction of the respective blue and green emission values of the stimulated cells. EC50 values
were calculated using ma 4-parameter non-linear analysis using Gen5 software (V2.03.1). The dose response
curves were drawn using Graph Pad Prism software (version 5.0).
4-(2-Hydroxy-2-phenylethoxy)-benzaldehyde 20: To styrene oxide (6.50g, 54.0 mmol) in toluene (85 mL)
was added 4-hydroxybenzaldehyde (9.89g, 81.0 mmol), PEG4000 (polyethylene glycol, 1.15g) and 1M aq.
NaOH (85 mL). The mixture was heated in an 80C oil bath, under nitrogen, overnight. After cooling to room
temperature the mixture was poured into EtOAc (250 mL), washed with brine (2 x 250 mL) and dried over
anhydrous sodium sulfate. Filtration and concentration in vacuo afforded a light yellow oil which was purified
by chromatography on silica gel (40 mm OD. 150g 230-400 mesh silica gel) using the flash technique, eluting
with 0-10% EtOAc / CH2Cl2. A higher Rf (TLC) fraction was isolated (20, 4.49g, 34%) and a lower Rf fraction
(19, 2.46g, 19%) was obtained. Compound 20 was determined to be the desired mode of addition and 19 the
regioisomer. Data for 20: 1H-NMR (400MHz, CDCl3);  = 9.89 (s, 1), 7.86 (d, J = 8.7 Hz, 2), 7.30 – 7.55 (5),
7.04 (d, H = 8.7 Hz, 2), 5.18 (m, 1), 4.80 (brs, 1), 4.02 – 4.21 (2); MS (ESI+) m/z 265.1 (M + Na).
5-[[4-(2-Hydroxy-2-phenylethoxy)phenyl]methylene]-2,4-thiazolidinedione: To a solution of 20 (2.63, 10.8
mmol) in absolute EtOH (75 mL) was added 2,4-thiazolidinedione (1.27g, 10.8 mmol) and piperidine (0.54 mL,
5.4 mmol). The resulting solution was heated to reflux under nitrogen for 10 hours. The mixture was cooled to
room temperature, acetic acid (20 drops) was added and the solvent was removed in vacuo to give a yellow oil.
The crude product was purified by chromatography on silica gel using the flash technique (40 mm OD, 150g,
230-400 mesh), eluted with 30-40% EtOAc-hexanes. Fractions containing the product were combined to furnish
3.18g (86%) of the Knoevenagel product as a pale yellow solid. 1H-NMR (400MHz, DMDO-d6);  = 12.53 (s,
1), 7.00-7.80 (9), 5.70 (m, 1), 4.80 (m, 1), 4.02 (m, 2); MS (ESI-) m/z 340.1 (M - H).
5-[[4-(2-Hydroxy-2-phenylethoxy)phenyl]methyl]-2,4-thiazolidinedione 21: To a solution of 5-[[4-(2-
hydroxy-2-phenylethoxy)phenyl]methylene]-2,4-thiazolidinedione (1.50g, 4.39 mmol) in THF (20 mL) was
added H2O (20 mL), 1M aq. NaOH (3 mL), cobalt (II) chloride hexahydrate (0.60mg, 0.003 mmol) and
dimethylglyoxime (15mg, 0.13 mmol). The resulting mixture was treated with a solution of NaBH4 (240mg,

6.33 mmol) in 0.2M aq. NaOH (3.6 mL) over 15 minutes to produce a deep blue-purple color. The pH of the
mixture was adjusted to ca. pH 10 during the course of the reaction by the addition of acetic acid. After the
reaction was judged to be complete (HPLC) the mixture was cast into EtOAc (250 mL), washed with water (2 x
250 mL), and dried (Na2SO4). Filtration and concentration in vacuo afforded crude 21 as a foamy, yellow solid.
The crude material was purified by chromatography on a column of silica gel using the flash technique (40 mm
OD, 150g, 230-400 mesh), eluting with 50% EtOAc-hexanes. Fractions containing the desired 21 were
combined and concentrated in vacuo to give 21 (1.15g, 76%) as an ivory solid. 1H-NMR (400MHz, CDCl3);  =
8.00 (brs, 1), 7.20-7.45 (5), 7.17 (d, J = 8.6 Hz, 2), 6.89 (d, J = 8.6 Hz, 2), 5.14 (m, 1), 4.52 (dd, J = 9.3, 4.0 Hz,
1), 3.95-4.10 (2), 3.46 (dd, J = 14.2, 3.9 Hz, 1), 3.14 (dd, J = 14.2, 9.3 Hz, 1), 2.78 (brs, 1); MS (ESI-) m/z 342.1
(M - H).
5-[[4-(2-Oxo-2-phenylethoxy)phenyl]methyl]-2,4-thiazolidinedione 8: To a stirring solution of 21 (1.00g,
2.91 mmol) in CH2Cl2 (35 mL), cooled in an ice-water bath, was added DMSO (2 mL). Phosphorus pentoxide
(0.83g, 2.91 mmol) was added followed by Et3N (1.8 mL, 13.1 mmol). The reaction was allowed to slowly
warm to room temperature and was analyzed by HPLC. When the reaction was judged to be complete the
mixture was cast into CH2Cl2 (250 mL), washed with water (250 mL), brine (250 mL) and dried (Na2SO4).
Filtration and concentration in vacuo afforded crude 8 as a yellow oil. The crude material was purified by
chromatography on a column of silica gel (40 mm OD, 150g, 230-400 mesh) using the flash technique, eluting
with 25-35% EtOAc-hexanes. Fractions containing 8 were combined, and concentrated in vacuo to furnish
0.40g (40%) of 8 as a white solid. 1H-NMR (400MHz, DMSO-d6);  = 12.03 (brs, 1), 8.02 (d, J = 7.2 Hz, 2),
7.70 (m, 1), 7.57 (m, 2), 7.15 (d, J = 8.6 Hz, 2), 6.91 (d, J = 8.6 Hz, 2), 5.56 (s, 2), 4.88 (m, 1), 3.30 (dd, J =
14.2, 4.2 Hz, 1), 3.05 (dd, J = 14.1, 9.3 Hz, 1); MS (ESI-) m/z 340.1 (M - H); Anal. Calcd. For C18H14NO4S: C,
63.33; H, 4.43; N, 4.10. Found: C, 63.01; H, 4.28; N, 4.32.
2-Bromo-1-(4-ethylphenyl)-ethanone: To a solution of 1-(4-theylphenyl)-ethanone (6.00g, 40.4 mmol) in
acetonitrile (200 mL) was added p-toluenesulfonic acid (10.46g, 60.8 mmol) followed by N-bromosuccinimide
(7.20g, 40.4 mmol). The mixture was warmed to reflux under nitrogen and maintained for 2 hours. The mixture
was cooled to room temperature, concentrated in vacuo, dissolved in CH2Cl2 (250 mL), washed with water (2 x

250 mL) and dried (Na2SO4). Concentration in vacuo afforded the crude bromide as a yellow-brown oil which
was purified by chromatography chromatography on a column of silica gel using the flash technique (40 mm
OD, 100g, 230-400 mesh) eluted with CH2Cl2. Fractions containing the target bromide were combined and
concentrated in vacuo to afford 8.54g (93%) of 2-bromo-1-(4-ethylphenyl)-ethanone as a pale yellow oil. 1H-
NMR (400 MHz, CDCl3): δ = 7.90 (d, J = 8.3 Hz, 2), 7.64 (d, J = 8.3 Hz, 2), 4.48 (s, 2), 2.74 (q, J = 7.6 Hz, 2),
1.27 (t, J = 7.6 Hz, 3).
4-[2-(4-Ethylphenyl)-2-oxoethoxy]benzaldehyde 22a: To a stirring solution of 2-bromo-1-(4-ethylphenyl)-
ethanone (2.85g, 12.5 mmol) in acetone (80 mL) was added 4-hydroxybenzaldehyde (1.76g, 14.4 mmol)
followed by potassium carbonate (1.99g, 14.4 mmol). The mixture was allowed to stir at room temperature at
which point HPLC analysis suggested the reaction was complete. The solvent was removed in vacuo and the
residue was partitioned between ethyl acetate (125 mL) and water (250 mL). The aqueous phase was extracted
with EtOAc (125 mL) and the combined organic layers were washed with brine (250 mL), dried over Na2SO4,
and concentrated in vacuo to afford crude 22a as a yellow oil. Crude 22a was purified by chromatography on a
column of silica gel using the flash technique (50 mm OD, 150g, 230-400 mesh), eluted with 5-10% EtOAc-
haxanes. Fractions containing 22a were combined and concentrated in vacuo to provide 2.07g (61%) of 22a as
an ivory solid. 1H-NMR (400 MHz, CDCl3): δ = 9.91 (s, 1), 7.94 (d, J = 8.3 Hz, 2), 7.86 (d, J = 8.7 Hz, 2), 7.36
(d, J = 8.3 Hz, 2), 7.05 (d, J = 8.7 Hz, 2), 5.40 (s, 2), 2.76 (q, J = 7.6 Hz, 2), 1.29 (t, J = 7.6 Hz, 3); MS (ESI+)
m/z 269.1 (M + H).
5-[[4-[2-(4-Ethylphenyl)-2-oxoethoxy]phenyl]methylene]-2,4-thiazolidinedione 23a: To a solution of 4-[2-
(4-ethylphenyl)-2-oxoethoxy]benzaldehyde 22a (1.58g, 5.91 mmol) in absolute EtOH (15 mL) was added 2,4-
thiazolidinedione (0.761g, 6.50 mmol) and piperidine (0.643 mL), 6.50 mmol). The resulting mixture was
warmed to reflux under nitrogen and maintained for 12 hours. The mixture was cooled to room temperature and
then in an ice water bath to furnish a precipitate. The solid was isolated by filtration, rinsed with ether, and dried
in vacuo to give 23a (1.21g, 56%) as a pale yellow solid. 1H-NMR (400MHz, DMSO-d6);  = 12.53 (brs, 1),
7.95 (d, J = 8.2 Hz, 2), 7.75 (s, 1), 7.55 (d, J = 8.7 Hz, 2), 7.42 (d, J = 8.2 Hz, 2), 7.12 (d, J = 8.7 Hz, 2), 5.68 (s,
2), 2.69 (q, J = 7.6 Hz, 2), 1.21 (q, J = 7.6 Hz, 3); MS (ESI+) m/z 368.1 (M + H); MS (ESI-) m/z 366.1 (M - H).

5-[[4-[2-(4-Ethylphenyl)-2-hydroxyethoxy]phenyl]methyl]-2,4-thiazolidinedione 24a: To a stirring solution
of CoCl2-6H2O (3.36 mg, 0.011 mmol) and 2,2’-bipyridine (70.7mg, 0.45 mmol) in water (16 mL) and THF (16
mL) was added 1.0N aq. NaOH (2 drops) followed by NaBH4 (284mg, 7.5 mmol). The resulting deep blue
mixture was cooled in an ice-water bath and a solution of 23a (0.70g, 1.90 mmol) in THF-DMF (2:1, 12 mL)
was added over 0.5 h. The reaction mixture was allowed to slowly warm to room temperature and stir for 18
hours. Acetic acid was slowly added until the pH of the mixture was ca. 6. The mixture was diluted with water
(25 mL) and extracted with CH2Cl2 (2 x 50 mL). The combined organic extracts were washed with brine (100
mL), dried (Na2SO4), filtered and concentrated in vacuo. The crude product was purified by chromatography on
a column of silica gel using the flash technique (20 mm OD, 30g, 230-400 mesh) eluting with EtOAc-CH2Cl2
(1:4). Fractions containing 24a were combined and evaporated in vacuo to give 0.54g (76%) of 24a as a white
solid. 1H-NMR (400 MHz, CDCl3): δ = 7.38 (d, J = 8.0 Hz, 2), 7.25 (d, J = 8.0 Hz, 2), 7.16 (d, J = 8.6 Hz, 2),
6.89 (d, J = 8.6 Hz, 2), 5.12 (dd, J = 8.7, 3.1 Hz, 1), 4.51 (m ,1), 3.92-4.09 (2), 3.46 (dd, J = 14.2, 3.9 Hz, 1),
3.13 (dd, J = 14.2, 9.4 Hz, 1), 2.75 (brs, 1), 2.67 (q, J = 7.6 Hz, 2), 1.27 (t, J = 7.6 Hz, 3); MS (ESI+) m/z 394.1
(M + Na); MS (ESI-) m/z 370.2 (M - H).
5-[[4-[2-(4-Ethylphenyl)-2-oxoethoxy]phenyl]methyl]-2,4-thiazolidinedione 5: To a stirring solution of P2O5
(0.50g, 1.8 mmol) in CH2Cl2 (8 mL), cooled in an ice-water bath under nitrogen, was added a solution of 24a
(0.33g, 0.89 mmol) in CH2Cl2 (8 mL) followed by DMSO (0.32 mL, 4.4 mmol). After stirring for 15 minutes i-
Pr2Net (0.46 mL, 2.7 mmol) was added. The mixture was allowed to stir for 1 h, then was cast into cold (ice-
water) saturated aq. NaHCO3 (50 mL) and extracted with EtOAc (2 x 50 mL). The combined organic layers
were washed with brine (100 mL), dried (Na2SO4), filtered and concentrated in vacuo to afford crude 5 as a
sticky yellow solid. The crude product was purified by chromatography on a column of silica gel using the flash
technique (20 mm OD, 30g, 230-400 mesh), eluting with 0-15% EtOAc-CH2Cl2. Fractions containing 5 were
combined and concentrated in vacuo to give 236mg (72%) of 5 as a white solid. 1H-NMR (400 MHz, CDCl3): δ
= 8.10 (brs, 1), 7.95 (d, J = 8.3 Hz, 2), 7.35 (d, J = 8.3 Hz, 2), 7.17 (d, J = 8.6 Hz, 2), 6.91 (d, J = 8.6 Hz, 2),
5.28 (s, 2), 4.51 (dd, J = 9.6, 3.9 Hz, 1), 3.48 (dd, J = 14.1, 3.9 Hz, 1), 3.12 (dd, J = 14.1, 9.6 Hz, 1), 2.75 (q, J =

7.6 Hz, 2), 1.29 (t, J = 7.6 Hz, 3); MS (ESI+) m/z 370.1 (M + H); MS (ESI-) m/z 368.1 (M - H); Anal. Calcd, for
C20H19NO4S: C, 65.02; H, 5.18; N, 3.79. Found: C, 65.13; H, 5.14; N, 3.91.
4-[2-(3-Methoxyphenyl)-2-oxoethoxy]benzaldehyde 22b: To a solution of 2-bromo-1-(3-methoxyphenyl)-
ethanone 14.85g, 64.83 mmol, Aldrich) in acetone (100 mL) was added 4-hydroxybenzaldehyde (8.21g, 67.2
mmol) and potassium carbonate (9.29g, 67.2 mmol). The mixture was allowed to stir at room temperature under
nitrogen for 12 hours, then the solvent was removed in vacuo and the residue was partitioned between CH2Cl2
(500 mL) and water (500 mL). The organic phase was washed with brine (500 mL), dried (Na2SO4) and
concentrated in vacuo to give crude 22b as a sticky pale yellow solid. Crude 22b was purified by
chromatography on a column of silica gel using the flash technique (50 mm OD, 150g, 230-400 mesh), eluting
with CH2Cl2. Fractions containing the desired product were combined and concentrated in vacuo to afford 22b
(10.84g, 62%) as an off-white solid. 1H-NMR (400 MHz, DMSO-d6): δ = 9.88 (s, 1), 7.86 (d, J = 7.5 Hz, 2),
7.63 (m, 1), 7.51 (m, 2), 7.21 (m, 1), 7.17 (d, J = 7.5 Hz, 2), 5.76 (s, 2), 3.84 (s, 3); MS (ESI+) m/z 271.2 (M +
H).
5-[[4-[2-(3-Methoxyphenyl)-2-oxoethoxy]phenyl]methylene]-2,4-thiazolidinedione 23b: To 4-[2-(3-
methoxyphenyl)-2-oxoethoxy]benzaldehyde 22b (5.10g, 18.9 mmol) in absolute ethanol (50 mL) was added
2,4-thiazolidinedione (2.21g, 18.8 mmol) and piperidine (0.28 mL, 2.8 mmol). The mixture was heated to reflux
under nitrogen and maintained at reflux for 12 hours. The mixture was cooled to room temperature then was
chilled in an ice-water bath. The resulting precipitate was collected by filtration, washed with CH2Cl2 and dried
in vacuo to provide 23b (5.48g, 76%) as a fine ivory solid. 1H-NMR (400 MHz, DMSO-d6): δ = 12.53 (brs, 1),
7.76 (s, 1), 7.63 (m, 1), 7.48-7.56 (4), 7.28 (m, 1), 7.14 (d, J = 7.5 Hz, 2), 5.70 (s, 2), 3.85 (s, 3); MS (ESI-) m/z
368.1 (M - H).
5-[[4-[2-(4-Methoxyphenyl)-2-hydroxyethoxy]phenyl]methyl]-2,4-thiazolidinedione 24b: To a stirring
solution of CoCl2-6H2O (23.5 mg, 0.077 mmol) and 2,2’-bipyridine (494mg, 3.15 mmol) in water (100 mL) and
THF (100 mL) was added 1.0N aq. NaOH (10 drops) followed by NaBH4 (1.98g, 52.5 mmol). The resulting
deep blue mixture was cooled in an ice-water bath and a solution of 23b (5.48g, 14.8 mmol) in THF-DMF (2:1,
75 mL) was added over 0.5 h. The reaction mixture was allowed to slowly warm to room temperature and stir