Introduction

The discovery of the 3,2’-bisindole indirubin as an antileukaemic agent goes back to a traditional Chinese medicine preparation, consisting of 11 herbal constituents, named Danggui Longhui Wan. This preparation has been and still is used traditionally for the treatment of various chronic diseases (Tang and Eisenbrand 1992; Han 1994). Qing Dai was initially characterized as the active ingredient, which corresponds to the well-known blue dye indigo naturalis, a 2,2’-bisindole (Li 1987). Indigo naturalis is used in traditional Chinese medicine because of its hemostatic, antipyretic, antiinflammatory, sedative, antibacterial, and antiviral properties. Further investigations revealed that the antileukaemic principle was not indigo but a minor byproduct of indigo naturalis, the red-coloured 3,2’-isomer indirubin (Li 1987; Wu et al. 1980; Zheng et al. 1979a; Zheng et al. 1979b; Zhang 1983) (Fig. 1).

Fig. 1
figure 1

Structure of Indirubin and indigo

Full size image

In clinical trials on 314 patients suffering from chronic myelocytic and chronic granulocytic leukaemia, indirubin was given orally at a dosage of 150–450 mg per day. Complete remissions were observed in 82 cases (26%), partial remission in 105 cases (33%), and beneficial effects were found in 28% of the cases (87 patients). The treatment was well tolerated and did not induce major side effects (Gan et al. 1985).

Studies exploring the mechanism of action indicated that indirubin inhibited DNA and protein synthesis in cell-free systems and in various cell lines, in a cell-free assay (Zhang ZN et al. 1985) as well as in rats inoculated with Walker-256 sarcoma (Wu et al. 1980; Chang and But 1996; Du and Ceng 1981). Alterations were reported in membrane morphology of the surface and endoplasmic reticulum of white blood cells of patients treated with indirubin (Lee et al. 1979). Long-term studies in animals demonstrated absence of bone marrow toxicity and hematotoxicity (Ji et al. 1981; Wan et al. 1981). In a 6-month chronic toxicity study in dogs, reversible diarrhoea and slight hepatotoxicity were recorded at dosages up to 25 times those used for human therapy. No changes in haematopoiesis, electroencephalogram activity or renal functions were observed (Wan et al. 1981).

In studies aimed at elucidating the potential mechanisms of action of indirubins, it was discovered that these 3,2’-bisindols act as potent inhibitors of cyclin-dependent kinases (CDKs) (Hoessel et al. 1999). Like some other agents inhibiting cyclin-dependent kinases, these compounds were also found to be highly effective inhibitors of glycogen synthase kinase 3β (GSK3β). GSK3β is an essential element of the WNT-signalling pathway and is also involved in multiple physiological processes, including cell cycle regulation by controlling levels of cyclinD1 and β-catenin, dorsal-ventral patterning during development, as well as insulin action on glycogen synthesis (Willert and Nusse 1998; Diehl et al. 1998; Yost et al. 1996; He et al. 1995; Emily-Fenouil et al. 1998; Cohen 1999; Summers et al. 1999). Furthermore, GSK3β together with CDK5 is responsible for most of the abnormal hyperphosphorylation of the microtubule-binding protein tau observed in paired helical filaments, which are diagnostic for Alzheimer’s disease (Imahori and Uchida 1997; Mandelkow and Mandelkow 1998). Therefore, effective inhibition of GSK3β might open novel therapeutic fields, such as treatment of certain diseases of the central nervous system, e.g., Alzheimer’s .

CDKs as cell cycle regulators

Initiation and progression through the cell cycle is a process tightly controlled by internal and external signals (Morgan 1997). Growth factors provide early signals that trigger entry into the cell cycle (Fig. 2). Progression through the cell cycle is achieved by a complex orchestration of kinases and of activator and inhibitor proteins (Van den Heuvel and Harlow 1993; Weinberg 1995; Bernards 1997).

Fig. 2
figure 2

Cell cycle

Full size image

Cyclin-dependent kinases (CDKs) play a crucial role in initiating and coordinating the cell-division cycle (Morgan 1997). When activated by their binding partners, the cyclins, CDKs phosphorylate proteins on serine or threonine residues, utilizing adenosine triphosphate (ATP) as a phosphate donor (Van den Heuvel and Harlow 1993). Cyclins and their related CDK catalytic subunits combine to stoichiometric complexes, thus generating the respective CDK holoenzymes. These holoenzymes are further activated by phosphorylation of specific amino acid residues in the CDK catalytic subunit (Senderowicz and Sausville 2000). To date, nine different CDKs and at least 16 different cyclins have been described in mammalian cells.

Cells are triggered from the quiescent state (G0 phase of the cell cycle) into the cell cycle by appropriate mitogenic signals and pass the G1 phase where they prepare for DNA synthesis during the ensuing S-phase. This is primarily mediated by cyclin D/CDK4- and cyclinD/CDK6-complexes which hyperphosphorylate and thereby inactivate the retinoblastoma protein (pRb). Hypophosphorylated pRB forms a multiprotein complex with a spectrum of binding partners, such as Abelson kinase (cAbl), proteins bearing an LXCXE-motif such as histone deacetylase (HDAC), DNA polymerase δ and D-type cyclins, as well as transcription factors of the E2F family (Fig. 3a). Phosphorylations, mainly by CDK4/cyclinD-complexes, lead to pRb hyperphosphorylation which induces release of binding partners such as E2F transcription factors and HDAC from pRb. (Tamrakar et al. 2000; Harbour and Dean 2000) As a consequence, E2F-mediated transcription is activated, leading to synthesis of proteins necessary for S-phase-progression, such as cyclin E, cyclinA, dihydrofolate reductase (DHFR), CDK1, E2F-1 (Mueller and Helin 2000). To ensure that pRb is kept in its hyperphosphorylated state, a positive feedback loop exists, in which cyclinE is involved (Fig. 3b). This results in further phosphorylation of pRb in late G1, thus driving cells into S-phase. (Sherr and Roberts 1999)

Fig. 3a,b
figure 3

a Regulation of G1-S-transition of the cell cycle; b Positive feedback loop in the release of the E2F-transcriptions factors at the end of the G1-Phase

Full size image

For early S-phase progression, CDK2 needs to associate with cyclinE and later with cyclinA (Fig. 2). The further sequence of events is governed by CDK1/cyclinA complexes in G2 phase whereas transition to M phase is orchestrated by CDK1/cyclinB and CDK1/cyclinA which prepare the cell for division. (Sherr and Roberts 1999)

The CDK/cyclin complexes are negatively regulated by a stoichiometric combination with small inhibitory proteins, the so-called endogenous CDK inhibitors. Cells express at least two families of CDK inhibitors: the CIP/KIP-family (p21CIP/WAF1, p27KIP1, p57KIP2), and the INK4-family (p15INK4b, p16INK4a, p18INK4c, p19INK4d/ARF). Members of the CIP/KIP-family inhibit the activity of CDK2-, CDK4-, and CDK6-complexes, whereas members of the INK4-family exclusively interact with CDK4- and CDK6-complexes. (Carnero and Hannon 1997; Sherr and Roberts 1995; Sherr and Roberts 1999; LaBaer et al. 1997; Cheng et al. 1999; Toogood 2001) Correctness of cell cycle progression is controlled at three distinct checkpoints. The first is located at G1-S-transition, the second (DNA-damage-checkpoint) in late G2, and the third in metaphase during mitosis. The G1/S checkpoint controls that DNA damage is repaired before S phase begins, whereas the G2- and M-phase checkpoints ensure DNA and chromatid integrity and check the readiness of the cell to accomplish mitosis.

Many tumour cells have a deregulated cell cycle and also lack appropriate checkpoint control, and potentially all human tumours have deviations in cell cycle control (Draetta and Pagano 1996). Mutations and/or aberrant expression have, for instance, been identified for pRb and p53 in bronchial carcinoma, for cyclinD in lung, breast and liver cancer, for CDK4 in sarcoma and melanoma, for cyclinE in colon, prostate, gastric and breast cancer, and for the INK4- and CIP/KIP-family in lung cancer, leukaemia, and many others. Thus, the aberrant function or expression in malignant cells of many of the proteins that drive the cell cycle makes them especially promising targets for therapeutic intervention.

The effects of low molecular weight CDK inhibitors on the cell cycle and their value for potential anticancer treatment have been extensively studied (Malumbres and Barbacid 2001; Sielecki et al. 2000). Such substances exhibit properties rendering them attractive as potential anti-tumour agents. First, they are potent anti-proliferative agents, arresting cells in G1 or G2/M (Soni et al. 2001; Damiens et al. 2001). Second, they trigger apoptosis, alone or in combination with other treatments (Edamatsu et al. 2000). Thus, irrespective of whether kinase activity of CDKs is inhibited and/or their expression is downregulated, CDK inhibiting agents represent a very promising area for the discovery of novel anticancer agents. (Coleman et al. 1997; Lee and Adams 1995; Meijer 1996; Walker 1998; Garrett and Fattaey 1999; Toledo et al. 1999; Gray et al. 1999)

Small molecules found to inhibit CDKs

Screening for low molecular weight inhibitors of CDKs resulted in early identification of various compounds of natural origin. These include staurosporine and its mammalian metabolite UCN-01, flavopiridol, and several purine derivatives, such as olomoucine, roscovitine the purvanalols, and the azepinones such as kenpaullone or alsterpaullone (Fig. 4).

Fig. 4
figure 4

CDK inhibitors

Full size image

These inhibitors share a flat heterocyclic ring system that fits into the ATP binding site of the kinase catalytic site, competing with ATP. This mode of binding also applies to the bisindole indirubin which represents the parent structure of a series of novel analogues prepared by our group to achieve improved pharmacological properties.

Topology of the interaction of indirubins with CDK2

CDK2, consisting of an amino-terminal domain (~80 residues) and a carboxy-terminal domain (~210 residues), represents the minimal catalytic kinase core of the CDKs in general. The N-terminal domain is mainly formed from β-sheets, whereas the C-terminal domain predominantly consists of α-helices. (Hoessel et al. 1999; Davies et al. 2001).

Crystal structures of CDK2 in complex with two representative indirubin derivatives, indirubin-3’-oxime (E231) and indirubin-5-sulphonic acid (E226), have been published recently (Hoessel et al. 1999; Davies et al. 2001).

The interaction of the inhibitor with the ATP binding site is characterized by formation of three hydrogen bonds formed between the indirubin “facial side”, representing an NH-CO-NH hydrogen bond donor/acceptor/donor triplet, and the Glu81-Phe82-Leu83 complementary peptide backbone (Fig. 5). Both indirubin-3’-oxime and indirubin-5-sulphonic acid share a similar binding mode: N1 of indirubin donates one hydrogen bond to the backbone oxygen of Glu81 and the amino group of Leu83 donates one hydrogen bond to the indirubin-C2 oxygen. An additional hydrogen bond donated by N1’ of indirubin interacts with the backbone oxygen of Leu83. However, the latter is thought to contribute only a rather small proportion to the inhibitor’s overall binding affinity because of relatively unfavourable angular parameters of this interaction. (Davies et al. 2001)

Fig. 5
figure 5

Representation of the hydrogen bond formation between CDK2 and E231 (R1=H; R2=N-OH) and E226 (R1=SO3H; R2=O), respectively

Full size image

The bent shape of the indirubin “facial side” complements the shape of the ATP binding pocket (Fig. 6). In addition, hydrophobic interactions between the inhibitor’s planar ring system and the hydrophobic amino acid residues participating in the binding cleft substantially contribute to the binding. Moreover, it also has been shown that the tight binding of indirubin derivatives into the ATP binding site remains largely unaffected by cyclinA binding, to form the active CDK2/cyclinA complex, although binding of cyclinA induces substantial rearrangements of the binding site domains. (Davies et al. 2001)

Fig. 6
figure 6

E226 (yellow) and AMPPNP (β-γ-imidoadenosin-5’-triphosphat; non-hydrolizable form of ATP) (white) in the ATP binding pocket of CDK2 (Davies et al. 2001)

Full size image

Association of cyclinA with CDK2 leads to the enzymes glycine loop (residues12–20) moving away from the binding cleft, thus exposing the ribose binding site to solvent (Davies et al. 2001). Furthermore, this association leads to conformational changes of several protein domains, one of which is the rotation of the N-terminal relative to the C-terminal domain, which causes a widening of the catalytic cleft. The bound inhibitor follows the movement of the binding pocket, maintaining its tight interaction with the adenine binding pocket.

Positions of the indirubin scaffold that appear suitable for the introduction of substituents are indicated in Fig. 6 (red arrows). First, the 3’ position points into the region that is occupied by the ribose moiety of ATP. This should allow linking to 5- or 6-membered sugar molecules or other appropriate substituents that might improve water solubility.

Second, position 5 appears suitable for molecular variations within limited steric requirements. The environment around the α-phosphate group of ATP participates in charge interactions. (Davies et al. 2001) Thus, a sulphonate or similar group introduced into position 5 of indirubin, as exemplified with the 5-sulphonate E226, leads to markedly increased binding affinity, due to ionic interaction with the ω-amino-group of lysine 33. Thus, E226 closely mimics ATP’s interactions with the binding site.

Structure activity relationship

Table 1 gives a very limited selection of results obtained from our extended structure activity relationship studies (SAR). The parent compound, indirubin (E211) which has already successfully been tested in the clinic, exhibits relatively low inhibitory activity towards various CDKs. Introduction of a sulphonate group into position 5 (E226) very strongly increases inhibitory activity with an IC50 value toward CDK1/cyclinB of only 5 nM (Hoessel et al. 1999). Unfortunately, E226 was found not to penetrate cellular membranes to any measurable extent and therefore did not show comparable inhibitory effects towards CDK in cells, nor did it show any cell growth inhibition. When the sulphonate group was replaced by a N,N-dimethyl sulphonamide group (E233) the inhibitory potency towards isolated CDK1 and CDK2 remained extremely high. Still, the compound was found not to be taken up sufficiently into tumour cells, and therefore showed only poor growth inhibition.

Table 1 IC50 values of some indirubin-derivatives on isolated/recombinant CDK and in the NCI 60 cell panel. *Values of the NCI 60 cell panel are given as log10 of the mean growth inhibition (GI) values [Molar]. (n.d. not determined)
Full size table

With regard to the 3’ position of indirubin, introduction of an oxime group represents a very simple molecular modification. The oxime (E231) exhibits an approximately fivefold to 50-fold improved inhibitory activity towards various CDKs, as compared to indirubin (E211).

Attachment of glucose was achieved, as exemplified with E671, by an ethylene bridged oxime ether β-glucosidal bond. This compound (E671) showed significantly improved solubility as compared to the parent indirubin (E211) and to the corresponding oxime (E231). Table 1 demonstrates, in conjunction with Fig. 8, the potent CDK inhibitory potential of E671, both on recombinant CDK2/CycE as well as on CDK4/CycD in human large cell lung carcinoma cells LXFL-529L. Potent antitumor efficacy is proven by a mean GI50 of about 3 µM in the NCI 60 cell line panel. As anticipated, the ribose binding domain within the ATP binding site of CDK obviously is able to accommodate the glucose substituent of E671. Attachment of maltose instead of glucose resulted in complete loss of CDK inhibitory activity, indicating that a diglucoside moiety is no longer fitting into the binding site. These few examples from our extended and ongoing structure activity studies exemplify structural modifications of the basic bisindole template that might lead to promising novel candidates for clinical trials, which are expected to be far more active than the parent compound indirubin. We also are working on the development of compounds with a differential potency profile with respect to CDK-subtype inhibition.

Effects on the cell cycle

We and others have shown that indirubins cause cell cycle arrest in G1/S or G2/M in a wide spectrum of human tumour cells, such as HBL-100, MCF-7, CCL-39, PC-12, L1210, K-562, and HL-60 (Hoessel et al. 1999). It was invariably found that these compounds, especially when incubated at concentrations of ≥10  µM, induce arrest in G2/M as well as in G1/S, in most cases followed by induction of apoptosis (Marko et al. 2001).

When HBL-100 cells were treated with E231 (15 µM; 30 h), no alteration in mRNA and in protein levels (CyclinA, cyclinB1, and CDK1) were observed (Damiens et al. 2001). In contrast, in nocodazole synchronized MCF-7 cells incubated with E231 after nocodazole release, we observed time-dependent alterations in protein levels of CDK1, cyclinB, and CDK1 in complex with cyclinB. As shown in Fig. 7, CDK1 in complex with cyclin B transiently increases, followed by a rapid and significant decrease of the active complex. Concomitantly, a strong increase of total cyclinB and a significant reduction of total CDK1 protein level is observed. In contrast, the level of “orphan” cyclinB, i.e., not in complex with CDK1, remains elevated, reflecting the situation in G2/M-phase (Marko et al. 2001).

Fig. 7
figure 7

Total cellular content of cyclin B and CDK1 and of cyclin B in complex with CDK1 in MCF-7 cells, synchronized by treatment with nocodazole and incubated with E231 for the indicated times. (Western blot with monoclonal antibodies against human CDK1 or cyclin B, respectively. Detection of cyclin B in complex with CDK1 by IP with monoclonal antibodies against CDK1 followed by Western blot with monoclonal antibodies against cyclin B.)

Full size image

In addition to alterations in cellular levels of G2/M cell cycle regulators, another possible mechanism of interference with the cell cycle machinery in tumour cells is by direct inhibition of the kinase activity of CDK/cyclin complexes. Accordingly, in G2/M-arrested MCF-7 cells incubated with 5 µM E231 after nocodazole release, a significant and dose-dependent inhibition of CDK1 kinase activity was observed (Marko et al. 2001). Similar results were found in other cells (Damiens et al. 2001).

In addition to their potent inhibitory effectiveness towards CDK1/cyclinB, some indirubins are also potent inhibitors of CDK2/cyclinA and E, and of CDK4/cyclinD1. In consequence, pRb phosphorylation is expected to be blocked, resulting in blocked activation of E2F transcription factors and cell cycle arrest in G1/S. Accordingly, incubation of Jurkat cells with E231 for 30 h resulted in a reduced phosphorylation of the retinoblastoma protein at Ser807/811 (Hoessel et al. 1999). Similar supporting results were found in LXFL-529L cells incubated with E671 for 24 h (Fig. 8a and Fig. 8b). A significant and concentration-dependent decrease of hyperphosphorylation of pRb on three characteristic CDK4/cyclinD phosphorylation sites (Ser780, Ser 795, Ser807/811) is observed, concomitant with a decrease in total pRb protein (Eisenbrand et al. 2004). Phosphorylation of Ser780 results in release of E2F transcription factors and this is most markedly blocked at E671 concentrations >1 μM. Phosphorylation of Ser 807/811 releases c-abl from pRb, whereas phosphorylation of Ser795 is a prerequisite for entry into S-phase (Tamrakar et al. 2000; Connell-Crowley et al. 1997). Both these phosphorylations are effectively blocked as well. These findings are well in line with the observed induction of cell cycle arrest in G1/S and G2/M phases.

Fig. 8a,b
figure 8

a Influence of E671 on total pRb level as well as on the three CDK4/cyclinD phosphorylation sites Ser780, Ser795, and Ser807/811 after 24 h incubation of LXFL-529L cells; b pRb phosphorylation levels of cyclinD/CDK4 characteristic phosphorylation sites detected by Western blotting in LXFL-529L cells incubated 24 h with indicated concentrations of E671 compared to DMSO-control (C)

Full size image

Taken together, these results demonstrate that novel indirubin derivatives are highly promising candidates for clinical testing as antitumour agents and that these compounds share the inhibition of cellular CDKs as a common mechanism of action. A selection of novel indirubin derivatives has already been tested in vivo in human tumour xenograft models in nude mice and have shown high antitumour activity (Fiebig et al. 2001).

Summary

CDKs propagate the cell cycle in a highly coordinated way and are crucial components of orderly cell cycle progression. Proteins involved in the complex orchestration of the cell cycle often are mutated or aberrantly expressed in tumour cells and therefore are deemed to represent excellent targets for therapeutic intervention. The bisindole indirubin, discovered in a traditional Chinese medicine recipe used against chronic myeloic leukaemia, was found to be a potent inhibitor of CDKs. Inhibition of CDKs by indirubins leads to cell cycle arrest, preferentially in G1 and/or G2/M-phase, in most cases followed by induction of apoptosis.

Indirubins bind to the ATP binding site of CDKs in an ATP-competitive manner. Interactions that mainly contribute to the inhibitor’s binding are three hydrogen bonds formed between the NH-CO-NH hydrogen bond donor/acceptor motif of the “facial side” of indirubin and the polypeptide backbone amino residues Glu81, Phe82, and Leu83. In addition, hydrophobic interactions between indirubin’s aromatic scaffold and the enzyme’s binding topology are relevant. Extremely potent inhibitory effects on isolated CDK2 are brought about by a sulphonate group in position 5, interacting additionally with the amino group of the side chain of lysine 33, thus favouring ionic interactions. The lack of cellular uptake of this compound is unfavourable and triggered the search for analogues with improved cell membrane permeability. Introduction of an oxime group in position 3’ allowed the generation of β-glycosides linked through an ethylene ether bridge as a very promising approach into novel drugs with improved bioavailability.

Although these compounds inhibit preferentially CDKs, they have been found to interact also with few other kinases, such as glycogen synthase kinase 3ß (GSK3ß), or, sporadically, with some protein kinase C (PKC) isoforms, with protein kinase A (PKA), casein kinase and c-Src tyrosine kinase. In terms of IC50 values for kinase inhibition on isolated enzymes, however, CDK inhibition appears of major relevance for antitumour efficacy. We are confident that these highly promising novel anticancer compounds will soon reach the clinic.