A RSK kinase inhibitor reporting its selectivity in vivoMorten Frödin

The presence of 491 very similar protein kinases in humans has proven to be a serious obstacle to generating specific kinase inhibitors and assessing their selectivity. Now, rational drug design has led to the generation of an irreversible kinase inhibitor with built-in features for demonstrating specificity in cells.

In 2006, more than 6,000 publications described the use of inhibitors of protein kinases to unravel cellular regulation. How do we know that the inhibitors selectively targeted the kinase under study, without also affecting many other members of the family of ~500 highly similar kinases? The answer to this question in most of these studies is: we don’t! When kinase inhibi-tors are added to cells, there is no way of assess-ing the extent of off-target inhibition on a global level in vivo. On page 156 in this issue, Cohen et al.1 tackle this central problem by reporting a kinase inhibitor that is capable of revealing its exquisite specificity in cells.

Protein kinases are enzymes that catalyze the transfer of the γ-phosphate of ATP to serine, threonine or tyrosine residues in sub-strate proteins in order to alter their function. Protein phosphorylation is the most general regulatory mechanism in eukaryotic cells, controlling most basal and specialized cellular processes. Abnormal protein phosphorylation is implicated in the progression of human dis-eases, including cancer, chronic inflammation and diabetes. Consequently, vast resources are being spent on the development of small-molecule inhibitors of protein kinases for treat-ing disease, and a few have entered the clinic. Kinase inhibitors have also been widely used as research tools in cell biology for many years. Nevertheless, only a few highly specific kinase inhibitors have been described, because most of these compounds function as competitive inhibitors of ATP, and the ATP binding site is highly conserved in the 491 members of the main human protein kinase family.

Recently, Cohen et al.2 reported a structural bioinformatics approach to rationally design an irreversible inhibitor of a protein kinase, in this case RSK (90 kDa ribosomal S6 kinase). RSK is a family of four kinases that are activated by extracellular signals such as growth factors and are thought to regulate cellular division,

growth and survival3. RSK is composed of a C-terminal kinase domain (CTD), which func-tions to activate an N-terminal kinase domain (NTD), which phosphorylates the substrates of RSK. The inhibitor (called fmk) is an adenine analog that exploits two residues at defined positions within the ATP binding site of CTD as selectivity filters: a reactive cysteine that covalently binds a fluoromethylketone group and a threonine that accommodates a bulky p-tolyl group on the adenine analog. Given that among all human kinases, only RSK1, RSK2 and RSK4 contain the combination of a cyste-ine and a threonine at these specific positions in the ATP binding site, fmk was assumed to be very specific, although its selectivity was not extensively confirmed.

In this issue, Cohen et al.1 report the design of an fmk derivative, fmk propargylamine (fmk-pa), that contains an elegant built-in feature for assessment of selectivity in vivo (Fig. 1). In brief, the propargylamine functional group can be conjugated to azide-labeled reporter mol-ecules (for example, fluorescent probes) via the so-called click reaction4 in lysates of cells that have been treated with fmk-pa prior to lysis. When the authors subjected such cell lysates to SDS-PAGE analysis, only bands corresponding to RSKs were labeled at concentrations of fmk at which RSK had been maximally inhibited in the cells. Thus, fmk-pa indeed seems to be a very specific inhibitor, and is very potent as well, as half-maximal inhibition was observed at ~30 nM. The authors then used fmk-pa to provide evidence that certain extracellular signals, such as lipopolysaccharide in macro-phages, can activate RSK via an unidentified kinase in a CTD-independent manner. This observation raises the interesting question as to the identity of this kinase and how general the mechanism is. Finally, the authors used fmk-pa to show that RSK contributes to the phosphorylation of the ribosomal protein S6 at sites essential for stimulation of cell growth.

Until now, the most stringent approach for assessing the specificity of a kinase inhibitor has been testing the compound against a panel of 50–250 kinases in vitro, which has been very useful in confirming or disproving the specific-ity of many widely used kinase inhibitors5. The

Morten Frödin is in the Biotech Research and Innovation Centre and the Center for Epigenetics, Ole Maaloes Vej 5, DK-2200 Copenhagen N, Denmark.e-mail: morten.frodin@bric.dk

Clickable propargylamine

Clickable reporter molecule

Fluoro-methylketone

p-Tolylgroup

fmk-pa

fmk

Adenine

N

N NH

N

NH2

N

N N

NH2

O

F

HO

N

N N

NH2

O

F

NH

Figure 1 Rational design of a specific, irreversible inhibitor of the protein kinase RSK with a clickable tag for assessment of selectivity in living cells. Based on crystallographic knowledge of the precise location of ATP in protein kinases, Cohen et al.2 first introduced two groups at specific positions on an adenine ring analog, which allowed the resultant compound (fmk) to exploit a unique combination of two residues at specific positions in the ATP binding site of RSK: a threonine allowed access of a bulky p-tolyl group to a hydrophobic pocket, and a cysteine was covalently modified by a fluoromethylketone group. To enable detection of RSK (and possibly other kinases) modified by fmk in living cells, Cohen et al.1 next generated the derivative fmk-pa, which contains a propargylamine functional group. This group allows conjugation of azide-labeled reporter molecules to fmk-pa–labeled proteins via the so-called click reaction in lysates of cells treated with the inhibitor. Labeled proteins can be revealed by, for example, SDS-PAGE and detection of the reporter molecule.

138 VOLUME 3 NUMBER 3 MARCH 2007 NATURE CHEMICAL BIOLOGY

N E W S A N D V I E W S

fmk-pa and BI-D1870 each have their particular advantages. fmk-pa offers the possibility of in vivo assessment of specificity on a global scale. Furthermore, by virtue of its covalent binding, fmk-pa inhibits irreversibly, meaning that RSK can be inactivated by a brief exposure of cells to the compound, followed by a washout. Thus, during subsequent experimentation, only RSK1, RSK2 and RSK4 are likely to be inhibited. Finally, because the specificity determinants for fmk-pa in RSK are known, it was possible to generate RSK mutants resistant to fmk-pa (point mutants in either selectivity filter2). If expression of such mutants rescues a particular cellular process or phenotype inhibited by fmk-pa, an off-tar-get effect of fmk-pa has essentially been ruled out. BI-D1870, on the other hand, targets all four RSKs and directly inhibits NTD, thereby also blocking CTD-independent activation of RSK. These properties probably account for the finding that in certain side-by-side compari-sons, BI-D1870 seems somewhat more efficient than fmk. In ideal experiments, both fmk and BI-D1870 should be used, because this would provide strong pharmacological evidence that is obtainable with very few kinases.

Of broader significance, the approach devel-oped by Cohen et al.1 is likely to encourage

other investigators to rationally develop spe-cific, irreversible inhibitors of other kinases with built-in features for selectivity assessment in vivo. A clickable functional group can proba-bly be incorporated into most kinase inhibitors with relative ease. It will be more challenging to incorporate a functional group positioned properly for covalent modification of a cysteine within or near the ATP binding site. However, the mounting number of published kinase–kinase inhibitor cocrystals is providing struc-tural information that will facilitate this task. A lot of kinases, cell biologists and patients are eagerly awaiting the future.

COMPETING INTERESTS STATEMENTThe author declares that he has no competing financial interests.

1. Cohen, M.S., Hadjivassiliou, H. & Taunton, J. Nat. Chem. Biol. 3, 156–160 (2007).

2. Cohen, M.S., Zhang, C., Shokat, K.M. & Taunton, J. Science 308, 1318–1321 (2005).

3. Hauge, C. & Frodin, M. J. Cell Sci. 119, 3021–3023 (2006).

4. Rostovtsev, V.V., Green, L.G., Fokin, V.V. & Sharpless, K.B. Angew. Chem. Int. Ed Engl. 41, 2596–2599 (2002).

5. Davies, S.P., Reddy, H., Caivano, M. & Cohen, P. Biochem. J. 351, 95–105 (2000).

6. Sapkota, G.P. et al. Biochem. J. 401, 29–38 (2007).7. Smith, J.A. et al. Cancer Res. 65, 1027–1034

(2005).

new approach of Cohen et al.1 offers the advan-tage of assessing selectivity against all kinases (and other proteins for that matter) expressed in the cell type under study. Moreover, because selectivity is assessed in vivo, the approach tests for inhibition of the several modification and conformation states that many kinases can adopt in cells. The various states may differ greatly in their susceptibility to a given inhibi-tor, as exemplified by the widely used MEK1 inhibitors U0126 and PD 98059, which target inactive but not active MEK1, the form present in most kinase panels5.

fmk-pa should greatly assist the identification of bona fide cellular roles of RSK—an investiga-tion that has been seriously hampered by the lack of specific inhibitors. In fact, the RSK field is privileged, because this year another excellent inhibitor, BI-D1870, was reported by Sapkota et al.6. BI-D1870 is a reversible inhibitor that shows remarkable specificity: in a panel of 54 protein kinases, only RSK1, RSK2, RSK3 and RSK4 were inhibited with high potency, and one or two other kinases were inhibited with relatively low potency. One other RSK inhibitor, SL0101, has been reported7, but its specificity in vitro and in vivo is unknown, because it has been tested against only a handful of other kinases.

Ask your friends about their bedtimes and you will be amazed at how different the answers can be. Though most people’s sleep-wake sched-ule follows the solar cycle, some individuals are ‘larks’ whereas others are ‘night owls’. From these tendencies, it seems that the internal body clock (circadian clock) is not running at the same speed in these people, thereby leading to altered clock phase. However, the mechanism for this change in speed or period length of a circadian cycle remains a mystery. Evidence for the involvement of specific genes that determine cycle length has come from the examination of lark families with early-morning awakening and sleep times (a short

period), a condition termed familial advanced sleep phase syndrome (FASPS). The cause for this syndrome is a mutation (S662G) in the gene encoding period 2 (Per2), which leads to hypophosphorylation of the Per2 protein1. The same phenotype is caused by a mutation in either casein kinase Iε (CKIε) or CKIδ, which normally bind Per2 at phosphorylated Ser662 and then phosphorylate adjacent serine resi-dues2. A new study now provides the missing link in this process by demonstrating that alter-ing the FASPS site in the Per2 protein leads to a change in its phosphorylation state3. The study shows that this site is critical for regulating Per2 concentrations and leads to either a faster or slower circadian clock.

Whereas circadian period length in Drosophila melanogaster is directly connected to Per protein phosphorylation and stability4, the molecular correlate for period length in mammals is not known. However, the clock

mechanism is similar in the two organisms. At the core of the mammalian molecular clock are two transcriptional activators (Clock and Bmal1) and two factors that act negatively on them (Per and Cry). The heterodimer of Clock-Bmal1 binds to the Per and Cry promoters, leading to their expression. The proteins enter the nucleus and disrupt Clock-Bmal1 action, thereby suspending the transcription of their own genes. Although it is not understood how Per2 functions in this mechanism, it seems to be involved in modulating transcription indi-rectly (it cannot bind to DNA). Because the above-described mechanism does not exactly run on a 24-h schedule, it has to be adjusted daily by light to keep the clock in tune with the environment.

Given that FASPS is caused by hypophos-phorylation of Per2, Xu et al.3 imagined that there would be an important relationship between Per2 phosphorylation and circadian

Urs Albrecht is in the Department of Medicine, Division of Biochemistry, University of Fribourg, Rue du Musée 5, 1700 Fribourg, Switzerland.e-mail: urs.albrecht@unifr.ch

Per2 has time on its sideUrs Albrecht

The circadian clock runs with a period of about 24 h and therefore allows mammals to predict sunrise at the cellular level. Phosphorylation of the clock protein period 2 influences this process by varying the clock’s period length.

NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 3 MARCH 2007 139

N E W S A N D V I E W S