Cellzome AG and Exelixis Inc. have independently designed the first highly selective inhibitors of the g-isoform of phosphoinositide 3-kinase.1,2 Exelixis plans to out-license its inhibitors, whereas last week's acquisition of Cellzome by GlaxoSmithKline plc gives the pharma a new class of compounds for inflammatory and autoimmune diseases.

Phosphoinositide 3-kinase (PI3K) plays a central role in signaling pathways that contribute to cell growth, proliferation, motility and survival. The kinase occurs in four isoforms. The a- and b-isoforms are expressed in many tissues and cell types, and the g- and d-isoforms are expressed in immune cells.

Aberrant activation of PI3Kα and PI3Kb is a key driver of many solid cancers, whereas activation of PI3Kg and PI3Kd can lead to hematological malignancies and inflammatory diseases.

For the past decade, the challenge has been to design PI3K inhibitors that selectively hit the isoform implicated in a given disease while sparing the function of the other isoforms. That has generally required chemists to optimize inhibitors that are 10-1,000 times more selective for one isoform over the others.3

Selective inhibitors of PI3Ka and PI3Kd have moved into the clinic, and inhibitors of PI3Kb are in preclinical testing ( "PI3Kg and PI3Kd inhibitor pipeline" Table 1, "PI3Kg and PI3Kd inhibitor pipeline"). Inhibiting PI3Kg has been more difficult.

Genetic data from a variety of animal models suggest selectively inhibiting PI3Kg could have a broad anti-inflammatory effect with potential utility in a range of diseases,4 including rheumatoid arthritis (RA),5 atherosclerosis6 and diabetes.7,8 Highly selective inhibition of PI3Kg also should avoid the cardiotoxicity that has been associated with some PI3K inhibitors in mice that antagonize the PI3Ka isoform (see Box 1, "Heartless PI3K inhibition").

Chemical scaffolds that work for the other three isoforms "often fail to potently inhibit PI3Kg," Christian Rommel told SciBX. "The ATP-binding pocket of PI3Kg is structurally distinct [from the other isoforms] and is more tight and narrow as well as less flexible."

Rommel is the former CSO of Intellikine Inc., which was acquired by Takeda Pharmaceutical Co. Ltd. this year for $190 million up front and up to $120 million in milestones. While at Intellikine, Rommel oversaw the development of INK1117, a PI3Ka-selective PI3K inhibitor that is in Phase I testing to treat solid tumors, and IPI-145, a dual inhibitor of PI3Kg and PI3Kd that is in Phase I testing to treat hematological malignancies.

Compared with dual inhibition of PI3Kg and PI3Kd, inhibiting only PI3Kg "is thought to be more suitable for myeloid-triggered inflammatory processes implicated in atherosclerosis and certain forms of metabolic disorders," said Rommel.

Going native

Cellzome researchers reasoned they might have a better chance of identifying a PI3Kg-selective PI3K inhibitor by using initial compound screens in whole-cell extracts rather than in panels of purified recombinant kinases.

The goal was to test the activity of inhibitors in a setting that mimics physiology, VP of Research Operations Gitte Neubauer told SciBX. "The targets are full length, post-translationally modified, and their interactions with other proteins are largely preserved," she said.

"There has been a growing realization that recombinant protein kinases, which are often truncated and/or fusion proteins, do not exhibit the same activity and drug-binding properties as kinases in native cells and tissues," said Matthew Patricelli, director of technology at ActivX Biosciences Inc.

Proteome-wide analysis of compound selectivity against native kinases in the presence of cellular cofactors "is particularly useful to inform decisions, for example, in the lead optimization phase or in the clinical candidate selection process," said Henrik Daub, SVP of science and technology at Evotec AG.

ActivX's KiNativ platform and Evotec's Kinaxo platform both allow for proteome-wide kinase inhibitor profiling in cell and tissue extracts.9,10

To identify molecules that could serve as starting points for optimization of highly selective inhibitors of PI3Kg, Cellzome used its Kinobeads chemical proteomics platform.

The platform uses a resin matrix coated with immobilized analogs of small molecules that broadly bind kinases. The resin is incubated in a fresh cell extract along with a test compound or vehicle.

The arrangement sets up a competitive binding assay in which the test compound competes with the resin to bind PI3Kg in the cell extract. Hits are then identified by reduced binding of PI3Kg to the resin, which indicates the test compound binds PI3Kg more strongly than the resin and thus could be the starting point for inhibitor optimization.

Previously, the researchers had validated the platform by confirming its ability to predict the known targets of a panel of multikinase inhibitors, including the cancer drugs Gleevec imatinib from Novartis AG and Sprycel dasatinib from Bristol-Myers Squibb Co.11

In the new work, the team modified the resin to bind PI3Kg as well as potential off-target kinases PI3Kd, mammalian target of rapamycin (mTOR; FRAP; RAFT1) and DNA-dependent protein kinase (DNA-PK), and then they screened a library of 16,146 small molecules for selective inhibition of PI3Kg.

The lead hit of that screen, CZC19091, was potent and at least 20 times more selective for PI3Kg over the off-target kinases. However, the compound was poorly active in cellular assays and showed poor exposure in rats.

The researchers then made three chemical modifications to CZC19091, which led to the highly selective PI3Kg inhibitor CZC24832. The compound was more potent than the parent molecule. CZC24832 also was 100 times more selective for PI3Kg over PI3Ka and PI3Kd and 30 times more selective over PI3Kb. The compound showed strong activity in cellular assays and good exposure in rats.

In a mouse model of collagen-induced arthritis, CZC24832 decreased bone and cartilage destruction by 53% compared with vehicle control, confirming the compound's anti-inflammatory activity.

Finally, the team used the inhibitor to uncover a new role for PI3Kg in the regulation of proinflammatory T cells. In a panel of human primary cell co-cultures exposed to inflammatory stimuli, CZC24832 inhibited RAR-related orphan receptor C thymus-specific isoform (RORg2; RORgT) expression, blocking differentiation of naïve T cells into T helper type 17 (Th17) cells and reducing the number of IL-17-producing cells.

"Our results revealed a previously undescribed role of PI3Kg in the regulation of TH17 differentiation, supporting the involvement of PI3Kg in the control of both innate and adaptive immune mechanisms," the authors wrote in their paper in Nature Chemical Biology.

Although the lead PI3Kg inhibitor identified in the Nature Chemical Biology paper "is a very good and selective in vivo probe compound, it does not have ideal physical-chemical properties. Further optimization would be needed before it might be considered a clinical candidate," Neubauer said. She declined to provide additional details, and GSK did not respond to requests for comment.

For obesity-related diseases, "it will be important to evaluate the toxicity of CZC24832 following chronic, in vivo administration in the relevant animal models," said Giovanni Solinas, professor of physiology and medicine at the University of Fribourg.

Solinas and colleagues have shown that PI3Kg promotes obesity and insulin resistance in mice on a high-fat diet.7

Cellzome and GSK have been working together since 2008 to discover and develop kinase-targeted therapeutics using the Kinobeads platform, and GSK had an exclusive option to license any product candidates.

Neubauer declined to disclose if any compounds that have been discovered under that deal target PI3Kg or other PI3K isoforms.

Back to basics

Exelixis opted for a different approach than Cellzome and stuck with the standard structure-based design methods Exelixis used to discover and optimize selective inhibitors of PI3Ka, PI3Kb and PI3Kd.

"We had two key things working in our favor that made us believe we could use a traditional structure-based design approach" to discover PI3Kg inhibitors, Exelixis CSO and EVP of discovery research Peter Lamb told SciBX. "First, we have a very large in-house screening library of about 4.5 million compounds that includes a broad range of scaffolds and chemotypes. Second, we have X-ray crystal structures of PI3Kg bound to some of those chemotypes, which, we believed, could inform and guide our optimization of inhibitors that would be at least 10 times more selective for PI3Kg over the other isoforms."

An in vitro high throughput screen of the company's entire compound library revealed two hits that, with subsequent optimization, generated a series of about 30 small molecules that bound PI3Kg with varying levels of potency and selectivity.

The four most potent and selective inhibitors had PI3Kg IC50 values of 5, 8, 18 and 34 nM and were all 10-500 times more selective for PI3Kg over the other isoforms. The compounds also had good plasma exposure in mice and rats.

In a mouse model of inflammation-associated mast cell degranulation, all four compounds decreased degranulation compared with vehicle and did so at levels comparable to those seen in Pi3kg knockout mice.

Finally, in a mouse model of chemokine-induced proinflammatory neutrophil recruitment, the 34 nM inhibitor lowered recruitment by 50% compared with vehicle (p<0.05).

Results were published in the Journal of Medicinal Chemistry.

"Based on the data we've gathered so far, we believe PI3Kg inhibitors have broad-spectrum anti-inflammatory activity and could find therapeutic uses in autoimmune indications like rheumatoid arthritis, multiple sclerosis and lupus, as well as in airway inflammation diseases such as asthma and COPD [chronic obstructive pulmonary disease]," said Lamb.

Exelixis wants to partner its PI3Kg inhibitor program. "We are not actively working on this program now, as almost all of the company's resources are focused on our lead cancer compound, cabozantinib," he said.

Cabozantinib (XL184), an inhibitor of c-Met receptor tyrosine kinase (MET; HGFR) and VEGF signaling, is in a Phase III trial to treat medullary thyroid cancer and in two Phase II trials to treat a broad range of solid cancers. Next month the company will present data on the trials at the annual American Society of Clinical Oncology meeting in Chicago.

Exelixis also is interested in licensing its preclinical PI3Ka inhibitor program, said Lamb. Earlier this year, the company exclusively licensed its preclinical PI3Kd program to Merck & Co. Inc.

The PI3Kg inhibitors in both papers are covered by patents.

Fulmer, T. SciBX 5(21); doi:10.1038/scibx.2012.539
Published online May 24, 2012


1.   Bergamini, G. et al. Nat. Chem. Biol.; published online April 29, 2012; doi:10.1038/nchembio.957
Contact: Gitte Neubauer, Cellzome AG, Heidelberg, Germany
e-mail: gitte.neubauer@cellzome.com
Contact: Marcus Bantscheff, same affiliation as above
e-mail: marcus.bantscheff@cellzome.com

2.   Leahy, J.W. et al. J. Med. Chem.; published online May 1, 2012; doi:10.1021/jm300403a
Contact: Henry W.B. Johnson, Exelixis Inc., South San Francisco, Calif.
e-mail: hjohnson@exelixis.com

3.   Fulmer, T. BioCentury 20(9), A9-A13; Feb. 27, 2012

4.   Rückle, T. et al. Nat. Rev. Drug Discov. 5, 903-918 (2006)

5.   Rommel, C. et al. Nat. Rev. Immunol. 7, 191-201 (2007)

6.   Fougerat, A. et al. Circulation 117, 1310-1317 (2008)

7.   Becattini, B. et al. Proc. Natl. Acad. Sci. USA 108, E854-E863 (2011)

8.   Kobayashi, N. et al. Proc. Natl. Acad. Sci. USA 108, 5753-5758 (2011)

9.   Sharma, K. et al. Nat. Methods 6, 741-744 (2009)

10. Patricelli, M.P. et al. Chem. Biol. 18, 699-710 (2011)

11. Bantscheff, M. et al. Nat. Biotechnol. 25, 1035-1044 (2007)

12. Lu, Z. et al. Sci. Transl. Med.; published online April 25, 2012; doi:10.1126/scitranslmed.3003623
Contact: Ira S. Cohen, State University of New York at Stony Brook, Stony Brook, N.Y.
e-mail: ira.cohen@stonybrook.edu
Contact: Richard Z. Lin, same affiliation as above
e-mail: richard.lin@stonybrook.edu

13. Wander, S.A. et al. J. Clin. Invest. 121, 1231-1241 (2011)


      ActivX Biosciences Inc., La Jolla, Calif.

      Bristol-Myers Squibb Co. (NYSE:BMY), New York, N.Y.

      Cellzome AG, Heidelberg, Germany

      Evotec AG (Xetra:EVT), Hamburg, Germany

      Exelixis Inc. (NASDAQ:EXEL), South San Francisco, Calif.

      GlaxoSmithKline plc (LSE:GSK; NYSE:GSK), London, U.K.

      Merck & Co. Inc. (NYSE:MRK), Whitehouse Station, N.J.

      Novartis AG (NYSE:NVS; SIX:NOVN), Basel, Switzerland

      Pathway Therapeutics Inc., San Francisco, Calif.

      Roche (SIX:ROG; OTCQX:RHHBY), Basel, Switzerland

      Sanofi (Euronext:SAN; NYSE:SNY), Paris, France

      State University of New York at Stony Brook, Stony Brook, N.Y.

      Takeda Pharmaceutical Co. Ltd. (Tokyo:4502), Osaka, Japan

      University of Fribourg, Fribourg, Switzerland