Normal colon epithelial cells, in their way to malignancy, pass through a sequence of stages, which include the formation of aberrant crypt foci (ACF), subsequently the formation of adenoma and finally carcinoma [1]. During this process of carcinogenesis genetic lesions that characterize this transition accumulate [2]. In one of the pathways leading to colon carcinogenesis, an initial step consists of mutations on APC (adenomatous polyposis coli) gene at chromosome 5q21, the gene that is the cause of the hereditary polyposis syndrome familial adenomatous polyposis (FAP). These lesions are already present in ACF (Fig. 1A). For the next step that results in the formation of adenomas, activating mutations of the K-ras oncogene take place. The transition from adenoma to carcinoma requires the loss of function of tumour suppressor proteins Smad4 (alternatively called deleted in pancreatic cancer 4–DPC4) and p53 [3]. This is accomplished by mutations of one copy of the genes encoding these proteins and loss of heterozygosity (LOH) at their normal allelic loci at chromosomes 18q and 17p, respectively. This linear sequence of events causes inter-connected deregulation of cell homeostasis that culminate in the malignant phenotype by excess proliferation, inhibition of apoptosis and loss of normal anoikis (inter-cellular adhesion-dependent apoptosis).

In another pathway leading to colon carcinogenesis, genes involved in DNA mismatch repair such as MSH2 (human MutS homologue 2), MLH1 (human MutL homologue 1) and PMS2 are mutated (Fig. 1B). These mutations are the cause of the hereditary colorectal cancer syndrome HNPCC (hereditary non-polyposis colorectal cancer) and result in the microsatellite instability phenotype in which frequent frameshift mutations occur in DNA sequences characterized by nucleotide repeats throughout the genome [4, 5].

Mutations that are the cause of the two hereditary colorectal cancer syndromes FAP and HNPCC are present in almost 100% of sporadic cases (about 85% have a somatically mutant APC and 15% one of the HNPCC causing genes mutated).

In addition to genetic lesions epigenetic changes play a role in the pathogenesis of colorectal carcinoma [6, 7]. These changes consist of DNA methylation of promoter sequences that lead to transcription silencing and histone acetylation, methylation and phosphorylation [6]. DNA methylation takes place in promoter regions rich in cytidine-phosphoguanosine (CpG) nucleotide sequences called CpG islands and leads to a phenotype known as CIMP (CpG island methylation phenotype). CpG methylation is present early in the adenoma–carcinoma evolution (Fig. 1) and is associated more often with the microsatellite instability phenotype [8].

The ubiquitin–proteasome system (UPS) once believed to be a cellular waste disposal basket has recently been established as an important regulatory system that modulates virtually every cell function. Its involvement in processes paramount in the establishment of neoplasia such as the cell cycle and apoptosis as well as the finding that it regulates the key anti-apoptotic transcription factor NF-κB has formed the basis for the exploration of proteasome inhibition as an anti-neoplastic strategy. The proteasome inhibitor bortezomib is already used in the treatment of multiple myeloma and is investigated in other cancers such as colorectal carcinoma.

Cox-2 (Cyclooxygenase-2) is the inducible form of the enzyme of prostaglandins biosynthesis pathway that converts the lipid arachidonic acid to prostaglandin G2 (PGG2) and subsequently PGH2 with a peroxydase and a cyclooxygenase activity, respectively. In contrast to the ‘housekeeping’ Cox-1, which performs the same enzymatic activity and is expressed constitutively in most tissues, Cox-2 is not expressed at baseline but is induced by inflammation and in several types of cancer. In colorectal cancer it has been found to be up-regulated and evidence from in vitro models and in vivo knock-out and trans-genic animal models point to an important role of cox-2 in colorectal carcinogenesis and carcinogenesis in other locations [9]. In addition, epidemiologic studies of non-steroidal anti-inflammatory drugs (NSAIDs) have shown a positive preventive effect of Cox inhibition in both genetically predisposed people and in the general population [10–12]. NSAIDs have now been found to possess Cox-independent anti-proliferative effects in many cell types and it is possible that at least part of their cancer preventive and therapeutic effects seen in various models is Cox-independent. Nevertheless, this fact does not lessen their value as anti-neoplastic agents.

In the paragraphs that follow, pathways affected by the above genetic lesions playing important roles in colorectal carcinogenesis will be described. Afterwards the effects of Cox-2 enzyme in these critical pathways and the role of ubiquitin–proteasome system-mediated degradation in colon carcinogenesis will be discussed, putting forward a rational for a combination of Cox and proteasome inhibitors in the treatment of colorectal carcinoma.

APC is a regulator of the Wnt signalling pathway. Wnt (the name derived from the fly homologue Wingless and the first mammalian family member Wnt1 initially called int-1) is a soluble factor that ligates the cell surface receptor Frizzled in co-operation with the co-receptor LRP5 (low-density lipoprotein receptor-related protein 5). Human genome encodes for about 20 Wnt gene analogs [13]. After Wnt ligation Frizzled activates the protein dishevelled, which in cooperation with GSK3β binding protein (GBP) inhibits the kinase GSK3β (glycogen synthase kinase-3β) [14, 15] (Fig. 2). This serine–threonine kinase is part of a multi-protein complex together with APC, Axin, Conductin, β-catenin and casein kinase II (CKII). A functional APC helps maintain this complex in which GSK3β comes in close conduct and phosphorylates β-catenin either directly or through activation of CKII. Phosphorylated β-catenin interacts with the F box protein βTrCP (β-transducin repeat containing protein) part of a SCF type ubiquitin ligase and is ubiquitinated and proteasome degradated [16]. A mutant β-catenin in which serine at position 37 is changed to alanine (S37A mutant) and thus it cannot be phosphorylated at this position, cannot be ubiquitinated and degradated [17]. β-catenin ubiquitination is a reversible process as de-ubiquitinating enzymes can interact with β-catenin and prevent its degradation [18].

Activation of Frizzled by Wnt inhibits GSK3β and maintains β-catenin in the un-phosphorylated state, which allows its translocation to the nucleus [19].The same effect can be achieved by debilitating mutations of APC that prevent the formation of APC/GSK3β/axin/conductin/β-catenin complex [20]. In the cell nucleus β-catenin interacts with the transcription factor TCF4/LEF (T cell factor 4/lymphoid enhancer factor) displacing the inhibitory protein Groucho and the β-catenin-TCF4 complex on the DNA recruits the transcription machinery to initiate transcription from their target genes. Included among those genes are key players for cell proliferation, apoptosis and metastasis [21–34] (Table 1). Other genes such as ZO-1 [26] and Ephrin B [35] are down-regulated by β-catenin-TCF4. In addition the transcription complex up-regulates some proteins without inducing their transcription directly, probably through induction of another transcription factor or through protein stabilization. Among these proteins are u-PAR [23] and β-catenin's own E3 ligase, βTrCP that targets it for protein degradation in the protea-some [36]. Cyclin D1 may also belong to this category of indirectly up-regulated genes at least in some experimental systems [37]. Finally β-catenin can induce the transcription of genes in TCF4 independent manner, possibly by interaction with different transcription factors such as FoxO (Forkhead box O). An example of TCF4 independent β-catenin-induced gene is that encoding for ARF (alternative reading frame) a positive regulator of p53 [38].

APC has β-catenin transcription-independent proapoptotic activity [39] the lack of which may be important in colon carcinoma cells with APC mutations. APC promoted apoptosis is dependent on caspase-8 and the exogenous death receptor apoptosis pathway. Caspase-8 inhibitors impede APC initiated apoptosis in a xenopus egg extract system. Furthermore, APC interacts directly with the Rac-specific guanine nucleotide exchange factor Asef. Rac is involved in normal cell migration and in colon cancer with mutated APC the lack of APC-Asef physiologic interaction may play a role in accumulation of cells in intestinal crypts [40].

Besides Wnt-initiated signalling, GSK3β inhibition can be mediated by other kinases such as akt (down-stream of ras), ILK (integrin-linked kinase) and protein kinace Cβ (PKCβ), which as a result stimulate β-catenin/TCF4 transcription [22, 41, 42] (Fig. 3). In contrast an alternative inhibitory regulation of β-catenin not involving the ubiquitin–proteasome system is mediated by the calpain system. μ-calpain mediated β-catenin degradation may be particularly important in colorectal cancer where, due to APC mutations, pro-teasome-mediated regulation is defective [43].

Another function of β-catenin involves its association with the adherens junction protein E-cadherin (Fig. 3). In adherens junction, the main inter-cellular adhesion point, β-catenin serves as the bridge between E-cadherin and β-catenin and the actin cytoskeleton. An equilibrium exists in the two functions of non-ubiquitinated β-catenin, the transcription function after shunting into the nucleus and the cytoskeleton and adhesion function as part of the adherens junction. This equilibrium is regulated by tyrosine phosphorylation at tyrosine residue 654, which facilitates interaction with E-cadherin while phosphorylation facilitating targeting for ubiquitination and proteasome degradation occurs at serine residues 33 and 37 [44, 45].

A third function of β-catenin independent of the two others as transcription factor and adherens junction component, involves inhibition of NF-κB through direct interaction with this transcription factor [46] (Fig. 3). This effect may be the cause of GSK3β requirement for NF-κB activation [47].

Disabling mutations of APC is the most common activation event in the β-catenin pathway and is present not only in all hereditary polyposis coli syndrome patients, but also in about 80% of sporadic colorectal cancer patients. β-catenin transcription can be activated by other means such as mutations of β-catenin itself [48] or other components of the pathway. Nevertheless, β-catenin mutations are found more commonly in small adenomas in comparison with larger adenomas or carcinomas, a fact that may imply that different events disabling the pathway are not functionally equivalent [49]. Activation of β-catenin transcription by APC mutations or mutations of β-catenin itself is not functionally equivalent to an up-regulated Wnt-initiated signalling because Wnt activates, except for the canonical β-catenin pathway another pathway that through kinases Tak1 and Nlk (Nemo-like kinase) phosphorylates and inhibits TCF4 [50] finely tuning transcriptional activity under physiologic conditions.

Finally, in addition to TCF4 whose transcription function is activated when bound to β-catenin, β-catenin binds and activates another transcription factor, FoxO (Forkhead box O), which is activated in response to oxidative stress.This activation results in cell cycle arrest [51, 52].

From the four functions of β-catenin discussed above only the promotion of TCF4-dependent transcription is tumour promoting. The three others (interaction with e-cadherin, promotion of FoxO-dependent transcription and direct NF-κB inhibition) are tumour suppressive through metastases inhibition by stabilizing inter-cellular interactions, cell cycle arrest following oxidative stress and inhibition of pro-liferation and angiogenesis promoting NF-κB-dependent transcription. An additional point relevant to this discussion is that β-catenin has further anti-proliferative effects as it leads to an increase of p53 levels, although p53 is not a transcriptional target of β-catenin and its mRNA does not increase in these conditions [53]. In contrast, inhibition of proteasomal degradation of p53 is involved as proteasome inhibitors nullify β-catenin's influence. p53's activator ARF is induced by aberrantly activated β-catenin and is responsible for stabilization of p53 [54]. This may explain the fact that APC debilitating mutations are sufficient to initiate colon carcinogenesis by adenoma formation but for progression to carcinoma additional genetic lesions are required.

The oncogene K-ras has activating mutations, especially in codon 12, in about 40% to 50% of colorectal carcinomas [55, 56]. Multiple inter-connected path-ways start from K-ras. In a first pathway the activation of Raf kinase by K-ras leads to the activation of MAPK cascade [57]. In this cascade, Raf, a MAP kinase, activates MEK1 and MEK2, dual specificity kinases, which activate ERK1 and ERK2 (extracellular-signal regulated kinases 1 and 2) and JNK1 and JNK2 (c-jun N-terminal kinases 1 and 2), the final result being the activation of multiple transcription factors among which AP-1, the complex of c-jun and c-fos has a prominent role in the effects of the path-way in cell proliferation (Fig. 4). Other transcription factors activated by ERKs include elk1 and c-myc [58]. MAPK activation downstream of activated K-ras is also involved in up-regulation of multi-drug resistance-mediating p-glycoprotein, leading to cancer cell resistance to several commonly used anti-neoplastic drugs [59].

In another pathway, K-ras activates the serine/ threonine kinase PI3-K. PI3-K activates in turn the kinase akt (or PKB- Protein Kinase B). Akt phospho-rylates several substrates among which is kinase IKK leading finally to the activation of the transcription factor NF-κB through phosphorylation and protea-some degradation of NF-κB's inhibitor I-κB. NF-κB is, in most occasions, a potent survival promoting factor through induction of transcription of prosurvival genes (Fig. 4). In addition akt phosphorylates and inhibits the pro-apoptotic bcl-2 family member bad and the upstream caspase of the mitochondria-dependent apoptosis pathway, caspase-9 [60]. The kinase GSK-3β is also a substrate of akt and when phosphorylated, it is inhibited and allows the activation of β-catenin, as discussed in the section above. Other target substrates of akt include the transcription factors FKHR and FoxO [61], which when phosphorylated are inhibited, mdm2, which following phosphorylation translocates to the nucleus and inhibits transcription of p53 target genes and mTOR (mammalian target of rapamycin). mTOR is activated by akt and maintains translation initiating factor eIF4 in a functional state by inhibiting 4E-BP. This leads to translation of proteins such as cyclin D, HIF and ornithine decarboxylase [62]. In addition mTOR activates the kinase p70^S6K another activator of mRNA translation. PTEN (phosphatase and tensin at chromosome 10) an inhibitor of the PI3-K/akt pathway [63] is also a target gene, completing a negative feed-back loop [64].

Two other ras-initiated pathways involve the activation of Ral nucleotide exchange factor RalGDS, having effects in cytoskeleton remodelling, endocytosis, exocytosis and cell proliferation and the activation of phospholipase C (PLC). PLC acts through Inositol-1,4,5-triphosphate (Ins[1,4,5]P₃) production to activate Ins(1,4,5)P₃ receptors involved in calcium signalling [65] (Fig. 4).

The two first-described ras activated pathways (Raf/MAPK and PI3-K/akt) are very well studied in oncogenesis. Although Raf/MAPK is considered to play a role mainly in cell proliferation and PI3-K/akt in inhibition of apoptosis, this is an over-simplification and both path-ways have effects in both carcinogenic processes as well as in anoikis inhibition and tumour angiogenesis.

Pathways stemming from K-ras are interacting with each other and may even have opposite effects in the regulation of target proteins, the final outcome depending on other factors such as additional inputs and the duration of each stimulus. For example, the PI3-K/akt pathway activates NF-κB through phos-phorylation and proteasome degradation of I-κB. Akt also phosphorylates and inhibits GSK-3β kinase resulting in accumulation of β-catenin. This protein, except for its well-described functions as a transcription factor and in cell–cell adhesion through its interaction with e-cadherin, has now been found to interact and inhibit NF-κB [46]. The final result on NF-κB activation may, thus involve the strength and duration of each signal as well as input from other pathways acting simultaneously or in sequence.

On other occasions Ras-initiated pathways impact on a target protein in multiple levels towards the same result. This is true for example for cyclin D, which is stabilized when GSK-3β is inhibited by akt. β-catenin activation by GSK-3β inhibition leads to cyclin D gene transcription dependent on the action of β-catenin/TCF4 transcription complex. Cyclin D gene is also a target of ERK after activation of the Raf/MAPK pathway by Ras [57].

Ligation of TGF-β to its cell surface receptor complex TβRII and TβRI initiates a signal transduction path-way by activating the proteins Smad2 and Smad3 (R-Smads), which in conjunction with the co-activator Smad4 (also known as DPC4–deleted in pancreatic carcinoma 4) co-operate with other transcription factors for the transcription of target genes. Smad6 and Smad7 are inhibitory type Smads that inhibit R-Smad mediated transcription (Fig. 5). Seven different TβRIs and five TβRIIs exist in vertebrate cells and may associate with each other in multiple combinations [66]. Activation of the TβRII and TβRI receptors facilitates additionally an interaction with PI3-K leading to akt activation. Moreover, TβRII/TβRI stimulates phosphatase PP2A, an inhibitor of the kinase p70S6K. Ras may also be activated by TRs either directly or through PI3-K activation [67].

Further inter-relations exist between TGF-β and Ras pathways. Activation of the MAPK/ERK pathway by Ras leads to inhibition of the TGF-β/Smad path-way by proteasome degradation of Smad4 [68]. Jab1, Roc1 and Smurfs are the E3 ligases mediating degradation. Mutations of Smads have the same effect inactivating TGF-β signalling by their targeting to proteasome degradation [69]. Ras inhibits TGF-β signalling yet by another mechanism, the prevention of proteasome-mediated degradation of Smad co-repressor TGIF [70] (Fig. 5). A reciprocal relationship exists as Smad4 loss of function leads to hyperactivation of Ras/ERK activity [71]. The simultaneous activation of Smads and Ras/MAPK and Ras/PI3-K pathways by TGF-β may underlie the described effects of TGF-β signalling in both inhibiting and promoting carcinogenesis. It appears that when Smad signalling is intact and MAPK/ERK is not hyper-activated by other stimuli, tumour inhibiting effects pre-dominate whereas when Smad mutations debilitate signalling the epithelial to mesenchymal transforming effects are favoured [67]. The ratio of Smad3 to akt has a correlation with the final cell fate and specifically with whether the cell will undergo apoptosis in response to TGF-β signalling [72].

TGF-β pathway has opposite to the Ras/MAPK pathway effects in the regulation of the cell cycle in a non-transcriptional level through stabilization of the cyclin-dependent kinase inhibitor p27 [73].This effect may involve inhibition of the proteasome by TGF-β[74]. In contrast, Ras activation promotes p27 phosphorylation at threonine 187, an event leading to ubiquitination of p27 by E3 ubiquitin ligase SKP2 and proteasome degradation [73].

An interaction of the TGF-β/Smad pathway with the Wnt/β-catenin pathway exists in several levels. The protein axin, which constitutes part of the β-catenin/GSK-3β/APC/axin complex interacts with Smad3 and facilitates its activation [75] probably integrating signals of an adaptor function of axin for Smad3, which promotes interaction of Smad3 with the TβR complex. In the level of transcription a co-operation of Smads with β-catenin/TCF4 in the activation of target genes has been observed [76]. In vivo compound heterozygote mice for APC and Smad4 develop more malignant polyps than APC mutants alone [77]. Obviously interactions of the two systems may occur indirectly, for example, because of interactions of both with Ras initiated pathways.

In a proposed model of TGF-β target gene modification, the wide array of target genes is divided in three categories [78, 79]. Smad3 is the main positive transcription regulator of the pathway and induces the first category of immediate-early genes. ERK assists in this induction while Smad2 suppresses them. Two categories of genes whose modulation takes place later than immediate-early genes exist. The first termed intermediate-induced genes are positively regulated by Smad3 and inhibited by ERK and Smad2. The second category termed intermediate-repressed genes are repressed by all Smad3, ERK and Smad2. The complexity of TGF-β effects is further produced by the fact that the expression of as many as 4000 genes (or about 10% of the estimated human genome genes) is rapidly modified after TGF-β stimulation [78]. Among the modulated genes of TGF-β are the cyclin-dependent kinase inhibitors p15^INK4B, p21^CIP1 and p27^KIP and caspases, which are induced and c-myc, which is repressed. Nevertheless, this model is not covering all cases of gene regulation by the TGF and Ras pathways. For example, the expression of the enzyme furin, a convertase involved in cellular secretory pathways and notably in TGF-β's own secretion is positively regulated by both Smad2 and p42/p44 MAPK [80]. Clearly co-operation of TGF-β/Smad and Ras/MAPK/PI-3K pathways is more complex and cell context specific.

Several components of the TGF-β pathway are mutated in human colorectal cancer. Smad4 mutations are late events in colorectal carcinogenesis and are present in about 20% of cases [81]. TβRII is mutated more frequently in cases with microsatellite instability due to the presence of a polyadenine repeat in its coding sequence, which makes it prone to replication errors [82]. In microsatellite stable tumours TβRII mutations are less frequent [83]. Mutations in Smad2 are also rare (about 6% of cases) in human colorectal cancer [84].

p53 protein, a molecule identified more than 25 years ago as an interacting partner of SV40 large T antigen, remains one of the most studied proteins in cancer research and new elements of its regulation are continuously discovered and clarified.

p53 is a transcription factor that transcribes several genes involved in cell cycle control and apoptosis [85–88] (Table 2). p53 transcriptional activity is regulated by a complex system of interacting mechanisms. Central to this regulation is the inhibitor mdm2 (mouse double minute 2, also called hdm2 in humans), which is an E3 ubiquitin ligase and tags p53 for proteasomal degradation. Mdm2 is inhibited by the protein p14^ARF (p19 in mouse), which is encoded by a gene in the same locus on chromosome 9p (and with the same base sequence) as the cyclin dependent kinase inhibitor p16^INKA, but transcribed in a different reading frame (thus the designation ARF for Alternative Reading Frame). p14^ARF being a negative regulator of mdm2 is a positive regulator of p53's action (Fig. 6). Ubiquitination and degradation of p53 is mediated, in addition to mdm2, by at least three other E3 ligases, COP1, Pirh2 and ARF-BP1/Mule [89–93].

Serine and threonine phosphorylation of p53 by several kinases in different residues lead to activation of its transcriptional activity. In response to DNA damage the kinase ATM (ataxia telangiectasia mutated), both directly and indirectly through the activation of kinases chk1/2, phosphorylates and activates p53. Other activators of p53 include GSK-3β, p38 and JNK [94].

Another modulation of p53 activity takes place in the level of co-factor binding. A great array of co-factors can bind to DNA-bound p53 mainly depending on post-translational modifications that p53 has undergone. Given that these post-translational modifications depend on the stimulus that has activated p53, it is derived that different stimuli that activate p53 (e.g. UV light, oncogenic mutations) result in the transcription of a different set of genes. At least to a certain degree, the final decision of whether a cell will undergo apoptosis or cell cycle arrest after p53 activation, is taken by the selection of co-factors recruited. Recruitment of ASPP (apoptosis stimulating protein of p53) co-factors ASPP1 and ASPP2/53BP2/Bbp as well as of the p53 family members p63 and p73 and of the JMY co-factor, favours the induction of genes that lead to apoptosis. In the other hand, recruitment of p300 leads to cell cycle arrest [95]. The oncogene c-myc plays a role in the decision for the cell fate after p53 activation. When concomitantly activated, it inhibits cell cycle arrest, leading to massive apoptosis.

p53 binds as a tetramer to specific DNA sequences comprised by the base sequence RRRC(A/T)(T/A) GYYY (where R is a purine and Y is a pyrimidine) through its central DNA binding domain (amino-acids 102-292). Non-specific interactions of the carboxy-terminal part of p53, through three lysine residues to several DNA sites, are nullified when the residues are acetylated. This acetylation is brought about by p300, which in this way aids to direct p53 towards specific target genes. In addition p300 poly-ubiquitinates already mono-ubiquitinated molecules of p53 (after the action of mdm2), a modification that eventually leads to p53 degradation.

p53 action may result not only in gene transcription induction but also in transcription repression. Genes repressed by p53 include bcl2 and its anti-apoptotic family member bcl-xl and survivin.

Particularly relevant to colon cancer, p53 is a regulator of β-catenin stability through its target gene Siah-1 [96]. Siah-1 ubiquitinates β-catenin in an APC-dependent, GSK-3β-independent manner. Siah mediated ubiquitination links p53 with the hypoxia response given that PHD (Prolyl hydroxylating domain containing) enzymes that hydroxylate the transcription factor HIF in order to interact with VHL are Siah–1substrates [97].

p53 mutations in colorectal cancer occur in about half the cases as discovered in studies examining either p53 gene sequence or p53 stability by immunohistochemistry (stability considered an indication that p53 is mutated and its physiologic turn-over is inhibited mainly due to its inability to function as a transcription factor and transcribe mdm2) [98].

The overwhelming majority (about 95%) of p53 mutations occurs in the core DNA-binding domain of the molecule and about three fourths are single mis-sense mutations. This fact underlies the need for not only loss of transcription function but also of the presence of mutant protein, which interferes with the function of the normal product of the other allele. Mutations that would result in truncated p53 or total loss of the protein would not interfere with the function of the normal p53, and thus, would not be so efficient from an oncogenic point of view. The normal p53 allele, which is also stabilized as its transcriptional activity is reduced due to the presence of the mutated protein interference, may be able to maintain some low level but critical transcription of target genes till a loss of heterozygosity event in its locus deletes it and shut off all p53 activity. This happens as a late event in colorectal cancer pathogenesis, already after the transition from adenoma to carcinoma [99].

Ubiquitin is a 76 amino-acid protein, which was initially discovered in the mid-1980s as a signal for degradation of misfolded cellular proteins through recognition by the multi-protein protease complex termed proteasome. Soon it became evident that the proteasome was not only a waste basket for misfolded proteins but also a master cellular regulator for several cellular functions through degradation of nascent proteins [100]. These include transcription, cell cycle control, apoptosis, DNA repair, MHC I antigen presentation and stress response [101]. A complex mechanism of ubiquitinating enzymes executes ubiquitination of hundreds of proteins to be recognized by the proteasome and degraded. Ubiquitination is executed in three main steps facilitated by three different classes of enzymes [102]. Ubiquitin-activating enzymes (also known as E1) bind through a cysteine residue and activate ubiquitin in an ATP-dependent manner. Afterwards E1-bound ubiquitin is transferred to a cysteine residue of an E2 enzyme (or ubiquitin-conjugating enzyme). In the third step of the process, a ubiquitin ligase (E3 enzyme) attaches ubiquitin from E2 to a target protein substrate through the -amino group of a lysine residue of the target protein and the c-terminal glycine residue of ubiquitin. Other ubiquitin molecules can be added to the first attached through lysine residues, usually the K48 [103]. Linkage of ubiquitin to proteins by other lysine residues such as K29 and K63 do not result in recognition by the proteasome for degradation but are involved in other cellular functions such as DNA repair and plasma membrane proteins endocytosis [104, 105]. For a protein target to be recognized and degraded by the proteasome a ubiquitin chain of at least four ubiquitin molecules needs to be attached. The elongation of the ubiquitin chain is sometimes helped by a fourth type of protein factor termed E4 [106]. A growing number of enzymes that take part in ubiquitination is recognized, including at least 50 different E2 enzymes [105] and several E3 type enzymes belonging to three families: the HECT domain, the RING domain and the U-box containing family [107].

The ubiquitination of a protein is not a one way event but de-ubiquitination may take place and there are more than 70 de-ubiquitinating enzymes (DUBs) in the human cell [105, 108] as was discussed in the case of β-catenin whose degradation is prevented by the de-ubiquitinizing enzyme Fyn [18]. Another example of de-ubiquitinizing enzyme is HAUSP (Herpesvirus-associated ubiquitin specific protease), which antagonizes ubiquitination of p53 [109]. Even the proteasome complex possesses de-ubiquitinizing subunits that in contrast to other DUBs couple de-ubiquitination to degradation [110, 111] helping recycling ubiquitin during the proteasome proteolysis.

The 26S proteasome is a multi-unit 2.5 MDa protease complex with a cylindrical structure. It consists of two functional divisions. The cylinders of the 26S structure is cupped in the two ends by the 19S regulatory subunit, which functions as the ubiquitinated-substrate recognition subunit, unfolds the substrate and de-ubiquitinizes it. This function is performed by a ring of 6 AAA+ proteins (ATPases associated with various cellular activities), which unfold substrates in an ATP-dependent manner [112]. Then the unfolded protein is passed to the central part of the 26S protea-some made-up from the catalytic 20S core consisting of two heptameric α-subunits and two heptameric β-subunits, both laid in a ring configuration. β-subunits are located centrally and the two β-subunit rings in the periphery at the two ends of the 20S core cylinder [113]. In eukaryocytes each unit of the α- and β-heptamer is encoded by a different gene such as there are a total of 14 genes and there are 2 copies of each gene in each cell. The catalytic 20S proteasome possesses three distinct protease activities, chymotryptic-like, tryptic-like and peptidylglutamyl activity [114]. β-subunits β5, β2 and β1 harbour the chymotrypsin, trypsin and peptidyl-glutamyl activity of the protea-some, respectively [115]. The importance of the three protease activities of the proteasome varies depending on the substrate protein [116].

Several transcription factors, transcription factor regulators, kinases, phosphatases, kinase inhibitors and other proteins with important roles in cell growth, proliferation, apoptosis and homeostasis are substrates of the proteasome [20, 61, 90, 117–140] (Table 3).

The prototypic paradigm of a protein regulated by the ubiquitin/proteasome system is the transcription factor NF-κB. NF-κB is in reality a family of related proteins that include REL-A (p65), REL-B, c-REL, p50 and p52, which bind their cognate DNA sequences as dimers [141]. An inhibitor molecule, I-κB is bound to NF-κBα and keeps it transcriptionally inactive. Signals that lead to NF-κB activation activate the kinase IKK, which phosphorylates I-κBα. Phosphorylated I-κBα is ubiquitinated by βTrCP E3 ligase and recognized for proteasome degradation, releasing NF-κB to start transcription.

NF-κB proteins are an example of another mechanism of regulation by the proteasome, namely the cleavage of a precursor protein to generate a mature product. Proteasome substrates are, in this example, the proteins p100 and p105, which are cleaved by the proteasome to the mature p52 and p50 subunits, respectively. Although it is not entirely clear how the proteasome activity leads to only partial degradation of this precursor proteins, initial data point to a model according to which the three dimensional structure of the protein substrate allows only partial insertion into the proteasome lumen [142, 143].

NF-κB is regulated by the ubiquitin/proteasome system through an additional interaction that involves ubiquitination of the protein TRAF6 (TNFR-associated factor 6), a factor recruited to the receptor complex of IL-1R after ligation with IL-1. After this ubiquitination TRAF6 interacts with the adaptor protein TAB2 (TAK1 binding protein 2) and the kinase TAK1 (TGFβ activated kinase 1). TAK1 activation leads to the phosphorylation of IKK subunit IKKβ and activation of the kinase, which phosphorylates NF-κB inhibitor I-κB [144].

Ubiquitination and degradation of β-catenin and I-κB that leads to NF-κB activation is served by the same ubiquitin ligase, βTrCP [145]. The Wnt/β-catenin/TCF4 pathway increases levels of βTrCP by a mechanism not involving transcription of its gene targets but rather stabilization of the protein ligase itself [36]. This is a negative feedback mechanism for the β-catenin pathway but may at the same time induce the activity of NF-κB and may be one of the factors explaining the increased activity of NF-κB in colon cancer [146, 147]. I-κB presents an additional common regulatory mechanism with β-catenin being a substrate of μ-calpain [148]. μ-calpain degradation depends on the PEST (Proline-glutamate-serine-threonine) domain of I-κB and may theoretically be an additional point of inter-connected regulation of β-catenin and NF-κB in conditions where μ-calpain is saturated.

The critical role of the ubiquitin–proteasome system in regulation of both apoptosis and the cell cycle is worth further discussion. This role is expected from a review of the list of most studied proteasome substrate proteins. Direct and indirect proteasome influences in both critical processes have been revealed.

Two distinct pathways for apoptosis termed extrinsic and intrinsic have been described [149]. The extrinsic pathway starts from cell surface death receptors, which recruit and activate up-stream caspases and mainly caspase-8, which in their turn activate executioner caspases-3 or -7 [150]. The intrinsic pathway is triggered by mitochondrial perturbation, which results in release of cytochrome c, apaf-1 and Smac/Diablo. Inhibition of the apoptosis inhibitors (IAPs) is relieved and the up-stream caspase-9 is activated to activate again executioner caspases. The UPS is involved in the regulation of core apoptosis machinery. Pro-apoptotic bcl-2 family member Bax, Bad, Bid and Bik, as well as Smac/Diablo and IAPs are proteasome substrates. All members of the conserved family of IAPs possess BIR (bacilovirus inhibitor of apoptosis repeat) domains and in fact by definition any protein that contains a BIR domain belongs to the IAP family [151]. BIR domains represent a cysteine rich motif of about 65 amino-acids that mediates interaction of IAPs with caspases and leads to caspase inhibition. IAPs contain also a RING domain, which allows interaction with E2 ubiquitin conjugating enzymes, a function that characterizes the E3 enzyme activity of IAPs [152, 153]. Through this activity IAPs inhibit and promote degradation of Smac/Diablo but in addition their own auto-ubiquitination and proteasomal degradation. Both the IAP/Smac and IAP/caspases interactions involve BIR domains of IAPs and in the case of Smac a so-called IBM (IAP-binding motif) four amino acids domain and a homologous to IBM domain in the linker region in the case of caspases [154, 155]. Probably IAPs in this manner represent a safeguard mechanism that neutralizes accidental release of Smac/Diablo [156] and other promoters of apoptosis such as AIF (apoptosis inducing factor) and HtrA2/Omi [155, 157] from the mitochondria to the cytoplasm in non-apoptotic conditions. When apoptosis is triggered by external or internal stimuli this mechanism is overridden and activation of caspases cannot be prevented by IAPs [158]. An additional molecule that can interact and inhibit IAPs in a manner analogous to Smac/Diablo and HtrA2/Omi is the elongation factor GSPT1/eRF. This polypeptide can be processed to a fragment that is released from its initial site, the endopasmic reticulum-associated microsomes and bind IAPs in the cytoplasm inhibiting their interaction with caspases [159]. The IAPs/GSPT1 interaction associates apoptosis with the RNA translation machinery.

A reciprocal interaction by which caspases activation during apoptosis inactivates proteasome, has been described [160, 161]. Specifically the chymotrypsin-like and peptidyl-gloutamyl proteasomal activities are affected due to degradation of the 19S subunits Rpt5 and Rpn10 involved in substrate recognition and Rpn2 involved in holding together the lid and base components of the 19S proteasome [160, 161]. Indeed the fact that activated executioner caspases degrade protea-some components leading to proteasome inhibition points to a crucial role of proteasome in cell survival that needs to be circumvented in order for apoptosis to proceed [162]. This happens despite the role that the pro-teasome plays in degrading IAP member XIAP and cIAP1 after a RING domain reciprocal interaction [163].

Both p53 itself and its co-factor ASPP2/53BP2 are proteasome substrates [119] and given that these factors have a crucial role in inducing apoptosis, this constitutes an additional regulatory role of UPS in apoptosis [164, 165].

The role of the ubiquitin–proteasome pathway in cell cycle regulation has recently begun to emerge [166]. Two types of E3 ligases the APC/C (anaphase promoting complex/cyclosome) and the SCF (Skp1/Cullin/F-box protein) related type ligases are involved in cell cycle regulation [167]. During the metaphase of mitosis, sister chromatids remain attached in the centromere until all chromatids are connected with the centrosome through the mitotic spindle microtubules. When this event happens in the end of metaphase signals in the kinetochore (the site of centromere attachment to kinetochore micro-tubules), which up till then were inhibiting APC/C, are silenced. APC/C is activated and ubiquitinates the protein securin, an inhibitor of the protease separase for degradation. Separase is then activated and cleaves the proteins cohesins that hold sister chromatids together. Thus each sister chromatid is attracted towards the opposite centrosome and anaphase begins [168]. Cyclin B, considered the master regulator of mitosis, which composes together with cdc2 the mitosis-promoting factor (MPF) is also a substrate for ubiquitination by APC/C [169]. MPF phosphorylates mitotic regulating proteins in order for a cell to get through the G2 phase to mitosis. When mitosis is completed cyclin B is ubiquitinated by APC/C to be degraded by the proteasome and the cell exits mitosis to G1.

As evident from the above discussions, key components of all four major molecular pathways involved in the pathogenesis of colorectal carcinoma (β-catenin, Smad4, p53 and NF-κB downstream of Ras) are regulated by the ubiquitin–proteasome system. Pharmaceutical (or other) proteasome inhibition would be expected to have anti-oncogenic effects by stabilizing tumour suppressors Smad4 and p53 and inhibiting anti-apoptotic and drug resistance promoting NF-κB. On the other hand stabilization of anti-apoptotic β-catenin is already taking place in most colorectal cancer tissues irrespective of proteasome inhibition due to APC debilitating mutations. Moreover, as discussed in the relevant section, β-catenin up-regulation may have tumour-suppressing properties due to non-transcriptional functions. Despite the fact that proteasome has a variety of substrates both pro-apoptotic and pro-survival, its dysfunction or inhibition is expected to have profound effects in cell homeostasis and push the balance towards apoptosis especially in unstable neoplastic cells. This hypothesis is supported by the fact that the proteasome is inhibited by caspases, executing apoptosis.

p27 a cdk inhibitor is an important regulator of cell cycle and a proteasome substrate. It has been found to be down-regulated by increased proteasome activity in colorectal cancer [170]. Metastatic colorectal tumour tissue displays lower p27 immunostaining than corresponding primary tumours [171]. E3 ubiquitin ligase skp2 (S phase kinase-associated protein 2) and its co-factor cks1 (Cyclin kinase subunit 1), which are involved in p27 ubiquitination and subsequent degradation by the proteasome, are over-expressed in less differentiated colorectal tumours [172, 173]. Their level correlate with poor prognosis compared with tumours expressing low levels of these E3 ligases. Colorectal cancer tissues display, in addition to enhanced NF-κB expression, enhanced IKKα expression [174] a fact arguing for the importance of the pro-teasome in NF-κB up-regulation in these carcinomas. Hence, overall, proteasome inhibition by pharmaceutical agents has the potential of suppressing colorectal carcinogenesis [175].

The importance of cox-2 enzyme (Cyclo-oxygenase 2 or Prostaglandin endoperoxidase H synthase 2) in the pathogenesis of colorectal cancer became evident in a serendipitous manner when it was found that patients with hereditary polyposis coli treated with non-steroidal anti-inflammatory drugs had regression of colonic polyps [176]. Cox-2 is hyper-expressed in the majority of colon cancers as well as in a variety of other cancer locations such as head and neck, breast, cervix, bladder, gastric and else-where [177]. Colorectal carcinomas in other primates display also increased expression of Cox-2 [178]. Its gene is a transcriptional target of both β-catenin/TCF4 and NF-κB transcription factors as well as k-ras downstream transcriptional programs [179–182] (Fig. 7). Co-operation of ERK1 and ERK2, p38/MAPK and NF-κB is required for transcription of Cox-2 by protease activated receptors [183] and is involved in Cox-2 induction by interferon γ[184] and exogenous carcinogens [185] in various cell types. NF-κB and p38/MAPK but not ERK1/2 are involved in the induction of Cox-2 by interleukin-1 β[186]. Transcription factor AP1 stimulates Cox-2 transcription and its inhibition suppresses Cox-2 expression [187]. An IL-6 regulatory element binding NF-IL6 and a CRE (cAMP-response element) are both present in the promoter of cox-2 gene. The NF-IL6 site is also important for Cox-2 induction by Ets family transcription factor PEA3 [188]. An NFAT (nuclear factor of activated T cells) response element on Cox-2 promoter is responsible for Cox-2 induction in colorectal cancer cells [189]. Additionally a peroxisome proliferator response element (PPRE) is present on Cox-2 promoter [190] and is responsive to PPARα but not PPARγ stimulation [191]. Post-transcriptionally a role for translational silencer TIA-1 has been suggested [192, 193]. Cox-2 mRNA has an AU-rich element (ARE) in its 3'-untranslated region (3'-UTR) which binds TIA-1 and results in the inhibition of mRNA translation.

The cox-2 protein is the inducible form of the enzyme that catalyzes, by a cyclo-oxygenase and an endoperoxidase action, the conversion of arachidonic acid [5, 8, 11, 14 eicosatetraenoic acid) to prostaglandin G₂ (PGG₂) and PGH₂, which is finally converted by specific prostaglandin synthases to different prostaglandins (PGE₂, PGD₂, PGF₂, PGI₂ and TXA₂) [194] (Fig. 8). These lipid metabolites are transported outside the producing cell by the ATP binding cassette family transporter MRP4 (multi-drug resistance protein 4) [195] and, through ligation with their cognate receptors, have various pro-survival, proliferative and pro-angiogenic effects in an autocrine and paracrine mode of action [196, 197]. Prostanoid receptors belong to the rhodopsin-type super-family and have seven trans-membrane domains [198]. Prostaglandin amounts entering the circulation are rapidly removed by enzymatic inactivation during their first lung passage [199]. For example, PGE2 is the ligand for four different EP receptors (EP1 through 4), which signal through Ca⁺⁺ mobilization and cAMP. In addition EP4 activates the PI-3K/akt pathway [200]. PI-3K activation is dependent on transactivation of EGFR by EP receptor and results in activation of akt kinase [201, 202]. Nuclear receptor family transcription factor PPARβ/ lies downstream to PI3K/akt in this pathway and ApcMin mice lacking PPARβ/ display decreased polyps formation after PGE2 treatment [203]. Nevertheless, the role of PPARβ/ in colon carcinogenesis is still controversial and other investigators find a tumour suppressive effect [204, 205]. Furthermore, EP receptors, upon PGE2 binding, activate a cytoplasmic G-protein coupled receptor, Gs, which interacts with axin. Gs/axin complex prevents axin from interacting with APC/GSK-3β and thus β-catenin is stabilized by remaining non-ubiquitinated and is able to carry out its transcription activity [206, 207]. In this way a positive feed-forward loop is perpetuated in the cancer cell as β-catenin activity results in further EGFR transactivation [208].

PGE2 produced by colorectal tumour cells has tumour-promoting effects by affecting the immune system [209]. It is able to increase the activity of regulatory CD4⁺CD25⁺ T cells, which have suppressive properties against tumour specific cytotoxic T cells and it increases the expression of the regulatory T cell specific marker Foxp3. Cox-2 inhibitors reverse these effects in an in vivo mouse model [210].

The importance of signalling from EP₁ receptor has been described in a mouse EP₁-knockout model where a decrease of aberrant crypt foci (ACF) formation by 60% after azoxymethane treatment compared with control animals was observed [211]. In contrast knockouts of the EP₃ receptors did not display a decrease in ACF formation.

The other prostanoids (PGD₂, PGF₂, PGI₂, TXA₂) have a single cellular surface receptor each (DP, FP, IP, TP, respectively). Thromboxane TP receptor has been found to mediate the action of trefoil peptides (TFFs) family of three heat, acid and protease resistant peptides secreted in the GI tract in inflammatory conditions and mediating intestinal wound healing [212]. Cox-2 activation induced by TFFs through the phospholipase C pathway results in TXA2 production, which ligates its receptor TP.TP couples with the G-proteins Gaq, Ga12 and Ga13 and promotes cell invasion [213]. This Cox-2/TXA2 mediated interactions may be involved in the association of inflammation with carcinogenesis in colon.

Prostaglandins concentration is further increased in colorectal cancer because, in addition to Cox-2 up-regulation, decreased expression of 15-hydrox-yprostaglandin dehydrogenase (15PGDH) is observed. This is the enzyme that oxidizes the 15(S)-hydroxyl group of prostaglandins to inactive 15-keto metabolites and has been found to have decreased expression in colorectal cancer cell lines, adenomas in APC mice and in human colorectal cancer samples compared with adjacent normal colon epithelium [214]. 15PGDH is normally a TGF-β induced gene and as a result dysfunction of this pathway in colorectal cancer may underlie decreased 15PGDH expression [215].

Although cox-2 is regulated by a variety of transcription factors, the fact that it is found up-regulated even in early stages of colorectal carcinogenesis and that it is hyper-expressed in all cell lines with a mutant APC, underlines the importance of Wnt/β-catenin/TCF4 pathway in its regulation.

Cox-1, the constitutively expressed form of the enzyme that catalyzes conversion of arachidonic acid to PGH₂ has a high homology and remarkably similar tertiary structure with cox-2 but a different regulation of expression [216]. A role for cox-1 in the formation of polyps smaller than 1 mm in size in a mouse model of intestinal polyposis has been reported whereas cox-2 was induced in larger polyps [217]. This may be related to both the level of prostaglandin production and arachidonic acid depletion, which may be sufficient with the cox-1 action to sustain small sized polyps but needs the robust inducible cox-2 action in larger polyps.

Arachidonic acid is the substrate of Cox-2. Arachidonic acid utilization by cox-2 leads to a decrease in its level and in consequence a decrease in the level of the pro-apoptotic lipid ceramide and increase of anti-apoptotic bcl-2 inhibiting apoptosis by an additional molecular pathway [218]. A similar effect is observed with the action of the enzyme FACL4 (fatty acid CoA ligase 4) that depletes free arachidonic acid by catalyzing its conjugation with co-enzyme A (Fig. 9) and activation for esterification [219, 220]. The pro-apoptotic role of arachidonic acid inside the neoplastic cell is further highlighted by the fact that FACL4 is over-expressed in colon adenocarcinoma [221] and other carcinomas [222].

Unesterified arachidonic acid promotes production of pro-apoptotic ceramide by activating hydrolysis of sphin-gomyelin [223]. Besides apoptosis promotion, ceramide induces expression of cox-2 [224] in a homeostatic mechanism that will lead to utilization of arachidonic acid.

A pathway of arachidonic depletion is important for the neoplastic colorectal cell because free arachidonic production may be increased due to activation of phospholipase A2 (PLA2), the enzyme that performs the reverse to FACL4 function, by activated MAPK [225, 226]. Cox-2 inhibition has the potential of both blocking the generation of prostaglandins and promoting accumulation of unesterified arachidonic acid, actions that lead to cellular demise. Nevertheless, knocking out cytoplasmic PLA2 in APC mutated mice decreases the size [227] and number [228] of polyps in the small intestine despite the fact that it is predicted to deplete rather than increase free arachidonic acid. These results parallel the decreased polyps in double APC and Cox-2 knock out mice [229] and underscore the role of prostaglandins in colon carcinogenesis. The pro-apoptotic role of arachidonic acid is far from being nullified by these results and its increased levels which are more evident in conditions of cPLA2 activation and Cox-2 inhibition promote opening of the mitochondrial permeability transition pores and release of pro-apoptotic proteins [230]. Furthermore, a high Cox-2 and low cPLA2 phenotype has been found by immunohistochemistry in high-risk patients [231].

Lipoxygenase enzymes (LOX) play also a role in the metabolism of arachidonic acid as well as its dietary precursor, linoleic acid. 15-LOX-1 metabolizes mainly linoleic acid to 13-S-HODE (13-S-Hydroxy-octadecadienoic acid) while its homologue 15-LOX-2 metabolizes arachidonic acid to 15-S-

HETE (15-S-Hydroxy-eicosa-tetraenoic acid) [232]. Other LOXs having arachidonic acid as a substrate include 12-S-LOX, 12-R-LOX, 8-LOX and 5-LOX, the first enzyme of the pathway that leads to the production of leukotrienes. Additional LOXs may exist [233]. The role of LOXs in colon carcinogenesis is far from clear, despite their obvious role as enzymes involved in arachidonic metabolism. 15-LOX-1 is found to be hyper-expressed in colorectal cancer [234] and may play a role in the non-Cox mediated induction of apoptosis by NSAIDs [235, 236]. Nevertheless, decreased expression of the enzyme in colon cancer has been also reported [237]. The leukotriene synthesizing enzyme 5-LOX is implicated in cell proliferation [238, 239] and combined Cox-2 and 5-LOX inhibitors are interesting novel drugs with potential colon cancer therapeutic activity beyond that seen with single enzyme inhibition [240].

Aspirin (Acetyl–salicylic acid, ASA), NSAIDs and the newer Cox-2 specific inhibitors, the so-called coxibs, have carcinogenesis-suppressing effects in colorectal cancer cells by Cox-2 inhibition and inhibition of prostaglandins synthesis. Based on these properties they have been proposed as cancer therapeutic and preventive agents [241, 242]. In many instances, though, these drugs have been found to have anti-proliferative and pro-apoptotic effects even in cells that do not express Cox-2 [243–246]. Thus it has been deduced that other mechanisms are in place to mediate those effects. Some of these mechanisms have been revealed and involve different unrelated actions of Cox inhibitors in diverse pathways, some of which have been documented for several drugs and others only for specific NSAIDs or coxibs (Table 4).

A direct inhibition of IKK kinase is a first mechanism of Cox-independent proliferation suppression by ASA [247]. This inhibition suppresses NF-κB activity by preventing I-κB phosphorylation. Additionally NSAIDs inhibit, as discussed, IKK in a Cox-dependent way by inhibiting PGE2 production, which activates through its receptor EP4 the PI-3K/akt pathway.

Another cox-independent action of cox inhibitors that promotes apoptosis is the induction of the enzyme spermidine/spermine acyl-transferase (SSAT) the major catabolic enzyme of oncogenic polyamines [248]. Natural polyamines putrescine, spermidine and spermine promote cell proliferation by inducing c-myc, c-fos, c-jun [245] and several pro-teine kinases genes [249]. A direct interaction of polyamines with DNA has been described [250]. In contrast a decrease of polyamines by a specific inhibitor of their synthesis, DFMO (α-difluoromethy-lornithine) leads to cell growth inhibition and cell cycle arrest [251] and an increase of JunD mRNA and DNA binding activity [252]. JunD is a member of the AP-1 transcription factor complex, which in con-trast to other members, c-Jun and JunB, promotes cell cycle arrest. Thus, the induction of SSAT, which transforms spermidine and spermine to their N-acetyl derivatives by aspirin, sulindac sulfone and other NSAIDs promote apoptosis in colorectal cancer cells [253–255]. SSAT induction is mediated by PPARγ transcription factor ligated by sulindac [253]. SSAT gene has two PPRE (PPAR response elements) in its promoter, one of which is required for PPAR binding [256]. Several other NSAIDs such as indomethacin, ibuprofen and flufenamic acid have been found to be ligand activators of PPARγ[257]. This transcription factor inhibits colorectal cell proliferation directly [258] and indirectly by suppressing tumour angiogenesis [259].

Induction of the enzyme 15-LOX-1 mediates still another mechanism of Cox-independent anti-tumour effect of Cox inhibitors. This enzyme converts the dietary precursor of arachidonic acid, linoleic acid to 13-S-HODE [233]. 13-S-HODE was found to be increased in colorectal cancer cells after treatment with NSAIDs. Apoptosis induced in these cells after NSAIDs incubation was inhibited when 15-LOX-1 was also inhibited [235, 236]. Besides Cox and 15-LOX-1 a third enzyme involved in arachidonic acid metabolism, cPLA2 is affected by NSAIDs. The mRNA levels of this enzyme are decreased after aspirin treatment of both Cox-expressing and not expressing colon cancer cells [260]. Concomitantly, PGE2 levels are decreased in those cells.

Treatment of colorectal cancer cells with Cox inhibitors induces the TGFβ super family member NAG-1 (NSAID activated gene 1, also known with five alternative names; PTGF, PLAB, PDF, MIC-1 and HP00269) an event that results in apoptosis [261, 262]. Transfection of the cells with a NAG-1 containing plasmid induced also apoptosis. NAG-1 is a p53 target gene [263, 264] but its induction by NSAIDs is not p53-dependent, nor correlates with their Cox inhibition potency [261].

Sulindac and other NSAIDs inhibit phosphorylation and activation of the kinases ERK1 and ERK2 leading to inhibition of phosphorylation of the pro-apoptotic bcl2 family member Bad and in inhibition of the anti-apoptotic effects of ERKs [265]. Although this action of NSAIDs may be Cox-dependent through inhibition of PGE₂ initiated EGFR transactivation up-stream of ERKs, it has been observed in both cells expressing and not expressing Cox-2 [266] and thus it must be considered at least in part a Cox-independent pro-apoptotic mechanism of Cox inhibitors.

Rac1 a member of the Rho family of small GTPases related to Ras is induced after aspirin treatment of colon cancer cells [262]. Rac1 has a role in colon epithelium differentiation and it is expressed in colon epithelial cells at the villus tips.

Degradation of β-catenin has been observed after NSAIDs incubation in colorectal cancer cells cultures [267] and in vivo in Min mice [268] and may be an additional mechanism of Cox-independent apoptosis promotion. β-catenin degradation is APC-independent as it has been noticed in SW480 bi-allelic APC mutant cells. Other invastigators have found an inhibition of β-catenin-dependent transcription after aspirin and indomethacin treatment but at the same time an increase in phosphorylated β-catenin [269].

Prostaglandins need to be transported outside the producing cells to bind their cell membrane receptors and exert their functions. This function, as mentioned, is executed by the transporter MRP4 [195]. Several NSAIDs have been found to inhibit prostaglandin transport [270] outside the cell, an action that further potentiates the effect due to decreased prostaglandins production from Cox-2 inhibition. MRP4 inhibition is exhibited by NSAIDs in different degrees. Indomethacin, ibuprofen and keto-profen are more potent inhibitors while diclofenac and the Cox-2 specific inhibitors celecoxib and rofe-coxib are much less active [195].

Aspirin treatment of colorectal cancer cell lines produces an increase in the expression of DNA mismatch repair (MMR) proteins hMLH1, hPMS2 and MSH6 independently of Cox and results in apoptosis [271]. Cell cycle arrest and apoptosis was produced, though, at the same degree in MMR deficient cell, a fact that argues for other mechanisms being at work in NSAIDs-mediated apoptosis in these cells.

Increased angiogenesis and induction of VEGF-A has been described in several neoplastic tissues that display increased expression of Cox-2 and there is a correlation between the levels of the two proteins [272, 273]. As a result Cox inhibitors may reduce neo-plastic angiogenesis through a Cox-dependent mechanism. In addition, at least two Cox-independent mechanisms have been described for the anti-angiogenic effects of these agents. Treatment with indomethacin or the Cox-2 selective inhibitor NS-398 (N-(2-(cyclohexyloxy)-4-nitrophenyl)methane-sulfon-amide) increases the level of VHL (Von Hippel Lindau factor) and prevents hypoxia-induced decreases of its level. As a result transcription factor HIF-1α and target gene VEGF remain low even in hypoxic conditions and angiogenesis is reduced [274, 275]. Another mechanism of angiogenesis inhibition by NSAIDs involves enhanced degradation of transcription factor Sp1 and Sp4, which are involved in VEGF transcription [276]. Sp (specificity proteins) transcription factor family members initiate their transcription from GC-rich promoter regions in genes that include except VEGF, p27, cyclin D, E2F1 and TGF [277].

APC mutations in the majority of colorectal carcinomas lead to unregulated transcriptional activity of β-catenin and increased expression of cox-2 from an early phase of colon carcinogenesis. Cox-2 over-expression is evident not only in colon cancer but also in a variety of other cancers witnessing for its importance in the pathogenesis of neoplasia. Cox-2 activity leads to the production of prostaglandins with proliferative and anti-apoptotic properties. Moreover cox-2 activity depletes arachidonic acid, which is an apoptosis promoting lipid by triggering ceramide production from sphingomyelin [219].

The transcription factor NF-κB has a central role in favouring proliferation, inhibiting apoptosis and mediating chemotherapy resistance and is activated in colorectal cancer cells [146]. Combined treatment by a cox and proteasome inhibitor inhibits NF-κB activity through several mechanisms. Proteasome inhibition stabilizes the NF-κB inhibitor I-κB, a proteasome substrate. Moreover it stabilizes β-catenin, which through direct interaction with NF-κB inhibits its activity [46]. Cox inhibitors inhibit directly the kinase IKK, an effect that may mediate cox-independent anti-proliferative effects of Cox inhibitors [247]. Finally, cox inhibitors, by inhibiting PGE2 production block one of the path-ways of PI-3K/akt activation. Akt is a kinase activator of IKK and, thus, its inhibition prevents downstream NF-κB activation. Thus, a combined inhibition of cox-2 and the proteasome will result in a complete inhibition NF-κB in multiple levels.

A co-operation of cox and proteasome inhibition could take place in stabilizing CDK inhibitor p27. A decreased serine phosphorylation of p27 as a result of reduced akt activity due to reduced PGE2 production by cox inhibition will be associated with a reduced degradation of the residual ubiquitinated p27 due to proteasome inhibition.

Proteasome inhibition by itself stimulates release of arachidonic acid [278] and inhibits proteolysis-mediated arachidonic acid up-regulation [279]. It up-regulates also Cox-2 and stimulates PGE2 production [280]. Cox-2 up-regulation is not due to decreased protein destruction but to increased production resulting from increased gene transcription mediated by the transcription factor C/EBP (CCAAT/enhancer-binding protein δ) [281]. Increased C/EBPδ binding to cox-2 promoter recruits CBP (CREB-binding protein) and leads to increased H3 and H4 histone acetylation enhancing transcription. Enhanced C/EBPδ DNA binding and cox-2 transcription up-regulation after proteasome inhibition is dependent on the kinases p38, PI3K and PKC and inhibition of JNK kinase [282]. By these actions proteasome inhibition is predicted to inhibit apoptosis in cells where Cox-2 is active by further potentiating its activity and may not be by itself the optimal treatment for colon carcinoma as witnessed by an initial clinical trial [283] with mediocre results. Nevertheless, drug plasma levels unable to consistently inhibit the proteasome may have been an additional problem in this trial. In contrast, concomitant pharmaceutical inhibition of Cox has the potential to prevent this PGE2 production stimulation and gives additional clinical rational for the combination.

Another untoward effect of proteasome inhibition in the treatment of colorectal carcinoma may be prevented by Cox inhibitors in the level of proliferation-promoting polyamines. Levels of polyamines as well as levels of ornithine decarboxylase (ODC), the main enzyme of their production from their precursor ornithine, have been found elevated in colon carcinoma cells and the intestinal mucosa of Min mice [284, 285]. ODC is a gene target of c-myc and this fact may explain its up-regulation in cells with APC mutations [255] and activated K-ras [286]. ODC is also a substrate of the proteasome and proteasome's pharma-ceutical inhibition may further increase its levels in colorectal cells with the final result of increasing polyamine production. This untoward effect can be inhibited by simultaneous induction of the enzyme SSAT by Cox inhibitors as previously discussed [287].

As a result of anti-proliferative and pro-apoptotic co-operation in multiple levels as well as independent effects of each, combined cox and proteasome inhibition represents a particularly promising avenue to explore in the treatment of colorectal cancer. The concept of combined targeted therapy has been a promising one. It has been investigated using combinations of NSAIDs [288–291] or bortezomib [292, 293] with various other agents. Due to the fact that proteasome inhibition has a wide range of results stemming from the wide range of protein substrates, some pathways affected by its inhibition may promote instead of suppress carcinogenesis. In these instances concomitant inhibition of cox-2 is helping in restricting these effects and allows for a shift of the balance towards carcino-genesis suppression. Combined treatment has the advantage of allowing the use of lower doses of each drug that can be therapeutically attainable. With the maximal tolerated dose, for example, of bortezomib, the only proteasome inhibitor used currently clinically, proteasome is inhibited only at a level of 40% to 70% compared to baseline activity, leaving 60% to 30% of its activity unaffected [294–296]. At the doses used for anti-inflammatory and analgetic treatment Aspirin and other NSAIDs present a non-negligible percentage of side effects mainly gastrointestinal that may become severe or even life threatening. Hence an escalation of their dose for cancer treatment would be difficult. In contrast a combination of proteasome and Cox inhibitors in doses similar to the currently used for each drug alone would avoid adding side effects and would exploit at the same time the anti-neoplastic effects of each class of drugs. In preclinal in vitro models the combination of proteasome inhibitors lacta-cystin or bortezomib and Cox inhibitors aspirin or sulindac have been found to enhance the effects of each drug alone in colorectal cell lines [297, 298 and IA Voutsadakis et al.: unpublished data]. Moreover the combination of bortezomib with sulindac enhances the tumour suppression effect of each drug in a mouse colorectal cancer xenograft model [298]. Further studies will help exploiting the anti-carcinogenic properties of proteasome and Cox inhibition and bring their combination to the clinic for the benefit of colorectal cancer patients.