Cyclin D1 in non-small cell lung cancer: A key driver of malignant transformation

1. Physiological roles of cyclin D1 

The cyclins and their catalytic partners, the cyclin-dependent kinases (CDKs), were identified by groups led by Leland Hartwell, Paul Nurse and Timothy Hunt; discoveries that led to the Nobel Prize in Physiology/Medicine in 2001. In the early 1970s, Hartwell and his group in Seattle were working with budding yeast (S. cervisiae) and discovered that the CDC (cell division cycle) genes are specifically involved in cell cycle control [1]. One of these genes, CDC28, controls the first step in the progression through the G1-phase of the cell cycle. Hartwell also introduced the concept of cell cycle checkpoints; DNA damage induced arrests that allow time for DNA repair before continuation of the cycle. In the mid 1970s, Nurse in London isolated the gene cdc2 (cyclin-dependent kinase 2) in fission yeast (S. pombe) [2]. He showed that Cdc2 (the protein encoded by cdc2) is a protein-kinase that is able to phosphorylate specific serine and threonine residues of other proteins, and that the function of Cdc2 is identical to that of Cdc28 in controlling transition from G1 to S phase. In 1987, Nurse and his group isolated the first human cyclin-dependent kinase, CDK1 (the original name of CDK1 was Cdc2 in both vertebrates and fission yeast, and Cdc28 in budding yeast), which led in turn to the identification of the 10 human cyclin-dependent kinases known today [3]. Hunt's group in London discovered the first cyclin protein in eggs from a sea urchin (A. punctulata) [4]. Cyclin molecules are almost undetectable in unfertilized eggs, induced upon fertilization, and degraded periodically at each cell division, hence their name. This work led to the identification of several human cyclins which are named alphabetically and divided into four groups (G1, G1/S, S, and M cyclins) based on their behaviour in the cell cycle. This classification is useful but not universally applicable, because the same cyclin can have a different function or expression pattern in different cell types, and may contribute to the control of more than one cell cycle phase.

Cyclins act together with their CDK partners to drive cells from one phase of the cell cycle to the next [5]. The role of the cyclin molecule in this process is thought to be to direct the CDK activity onto appropriate target proteins. The CDK molecules have therefore been compared with an engine and the cyclins with a gear box, controlling whether the engine will run in the idling state or drive the cell forward in the cell cycle. In the G1 phase, D-type cyclins bind almost exclusively to CDK2, CDK4 and CDK6, with the cyclin D1/CDK4 complex having the highest kinase activity among these complexes. In S phase, cyclins A and E interact preferentially with CDK2, and in M phase, cyclins B and E interact preferentially with CDK1. The activity of CDK proteins is negatively controlled by several inhibitors. The Ink4 proteins p15INK4A, p16INK4B, p18INK4C, and p19INK4D form binary complexes with CDK4 and CDK6, thereby blocking the G1 to S phase transition, whereas the Cip/Kip proteins p21Cip1 and p27Kip1 bind to the entire cyclin/CDK holoenzymes, inhibiting kinase activity and transitions at all stages of the cell cycle [6].

Three D-type cyclins have been described in mammals so far: cyclin D1–D3. Cyclin D1 was reported independently by three different groups in 1991. Andrew Arnold's group in Boston identified a gene named PRAD1 (parathyroid adenomatosis 1) in parathyroid adenomas [7]. The group found that pericentromeric inversion of chromosome 11 is frequent in a subset of these benign tumours. The chromosome 11 inversion juxtaposes the promoter region of parathyroid hormone (PTH) to the PRAD1 gene, which leads to massive overexpression of PRAD1. It was found that PRAD1 encodes a protein of 295 amino acids with sequence similarities to the cyclins. The PRAD1 protein co-precipitated with cdc2 and induced phosphorylation of histone H1, a known substrate of cdc2. At the same time, Charles Sherr's group in Memphis stimulated murine macrophages with colony-stimulating factor 1 and identified three genes that were subsequently over-expressed, CYL1, CYL2 and CYL3 (cyclin-like genes 1–3) [8]. Expression of these genes was induced in the G1 phase of the cell cycle. The proteins were phosphorylated and associated with a protein similar to Cdc28/cdc2. Simultaneously, David Beach and colleagues at Cold Spring Harbor Laboratory in New York isolated a human gene that was named CCND1 (cyclin D1) from a glioblastoma cDNA library, based on the ability of this gene to compensate for loss of cyclin in a cyclin-deficient yeast strain [9]. Sequence comparison showed that CCND1 and PRAD1 were the same human gene, and that CYL1 was the murine homologue.

A large number of subsequent studies have reported on the physiological roles of cyclin D1 in normal cells. These analyses showed that the expression of cyclin D1 is induced by numerous mitogens, which are sensed by growth factor and cytokine receptors, leading to the activation of intracellular signalling pathways. For example, stimulation of the PI3K (phosphatidylinositol-3 kinase)/Akt pathway leads to activation of NF-kB (nuclear factor kappa B), a transcription factor that induces production of the cyclin D1 protein [10], [11]. Conversely, expression of cyclin D1 is repressed by other factors, including p53 [12]. As mentioned earlier, the first identified function of cyclin D1 is related to the control of G1–S phase progression (Fig. 1). During G1, following a mitogenic signal, cyclin D1 complexes preferentially with CDK4 but alternatively with CDK6 and CDK2, although, the cyclin D1/CDK2 complex is not active as a kinase. The p21Cip1 and p27Kip1 proteins act as assembly factors for cyclin D1 and its CDK partners. The cyclin D1/CDK4 complex is phosphorylated by CDK-activating kinase (CAK), and activated CDK4, targeted by cyclin D1, can hyperphosphorylate the retinoblastoma tumour suppressor protein (Rb1) [13]. Phosphorylation of Rb1 leads to dissociation of E2F (E2 promoter-binding protein dimerization partners) from the Rb1/E2F complex [14]. Dissociated E2F induces transcription of cyclin E and other genes which are required for entry into S phase. Additional substrates of the cyclin D1/CDK4,6 complex include Smad3, CDT1, p107 and p130, which all act in cell cycle control. Near the end of G1 phase, cyclin D1 is rapidly exported from the nucleus and degraded in the cytoplasm. Nuclear export requires phosphorylation of cyclin D1 at threonine (Thr) 286 by GSK-3β (glycogen synthase kinase 3β), which in turn allows Cmr1 nuclear exportin to shuttle cyclin D1 out of the nucleus [15], [16]. Following nuclear export, cyclin D1 is degraded by the proteasome, which requires ubiquitination by ubiquitin-ligases [16], [17]. However, the ubiquitin ligase that specifically recognizes cyclin D1 has not yet been identified. Under certain conditions, the cyclin D1/CDK4 complex can also act as a mediator of programmed cell death [18], [19].

View Large Image Download to PowerPoint

Fig. 1. Cyclin D1 function and localization. During the G1 phase, cyclin D1 assembles with CDK4 and a CIP/KIP protein. This complex enters the nucleus and phosphorylates Rb, which promotes release of E2F transcription factors and thus progression from G1 to S phase. During G1/S transition, GSK-3β enters the nucleus and phosphorylates cyclin D1 at Thr 286. This allows for cyclin D1 nuclear export by Crm1 and degradation by the 26S-proteasome. GSK-3β localization is controlled by FRAT/GBP. Several alterations can lead to cyclin D1 nuclear accumulation, including impaired export and degradation. Cyclin D1 can also act as a CDK-independent regulator of transcription (not shown here). Abbreviations: GSK-3β, glycogen synthase kinase 3β; Thr, threonine.

Cyclin D1 can also act independently of the CDKs, as demonstrated by several studies. Cyclin D1 directly interacts with and represses or induces several different nuclear receptors and transcription factors, including the estrogen receptor (ER), androgen receptor (AR), peroxisome proliferator activated receptor gamma (PPARγ), and signal transducer and activator of transcription 3 (STAT3) [20], [21], [22], [23], [24]. Knudsen and her group carried out pioneering work in this field and showed that cyclin D1 interacts with the androgen receptor and inhibits its transactivation potential, without affecting expression of the androgen receptor [24]. The transcriptional function of cyclin D1 links this molecule not only to the cell cycle and to apoptosis, but also to migration, invasion, differentiation, inflammation and angiogenesis.

2. Cyclin D1 in cancer 

The identification of cyclin D1 originated from the work with parathyroid adenoma and glioblastoma. Cyclin D1 was subsequently found to be overexpressed relative to normal tissue in a large spectrum of human cancers. Importantly, overexpression was often correlated with chromosomal translocation or increased gene copy number. Most notably, the t(11;14) (q13;q32) translocation, a characteristic of 70–90% of mantle cell lymphomas, was found to juxtapose the IgH heavy chain promoter with the CCND1 gene (also referred to as B-cell leukaemia/lymphoma 1 gene, bcl-1), resulting in constitutive overexpression of cyclin D1/CCND1 [25]. The t(11;14) translocation is also present in other haematological malignancies, including multiple myeloma, although at a much lower frequency [26]. Increased CCND1 gene copy number has been reported in breast cancer, oesophageal cancer, laryngeal cancer and NSCLC (see Section 4) and was generally associated in such studies with CCND1 overexpression [27], [28], [29], [30], [31], [32], [33]. At least two studies reported on a significant number of breast cancers with cyclin D1/CCND1 overexpression and normal gene copy number, pointing towards additional mechanisms leading to overexpression of cyclin D1 [30], [31]. Initial sequencing studies failed to reveal point mutations of CCND1 in breast cancers and parathyroid adenomas that were associated with cyclin D1 expression [32].

Although overexpression of cyclin D1 was found to be associated with many cancers, forced expression of cyclin D1 alone did not to induce malignant transformation in rodent fibroblasts or lymphocytes [34], [35]. Another study showed that the HER2 and HRAS oncogenes had the capability to activate the CCND1 promoter in breast epithelial cells, and ablation of cyclin D1 prevented oncogenic signalling [36]. These results led to the initial paradigm that cyclin D1 by itself may not act as an oncogene, but very recent data indicate that cyclin D1 can indeed cause malignancy if the protein accumulates in the nucleus. Diehl and his group have previously shown that the T286A mutant cyclin D1 is refractory to nuclear export and proteolytic degradation, because this mutant cannot bind to the Crm1 nuclear exportin [37]. In their most recent report, the group demonstrated that transgenic mice expressing the T286A mutant cyclin D1 develop mature B-cell lymphoma [38]. Expression of the T286A mutant cyclin D1/CCND1 in these mice resulted in S phase entry in resting lymphocytes, and the onset of B-cell lymphoma correlated with aberrations in the p53-pathway and overexpression of Bcl-2. Diehl and colleagues concluded from these important results that nuclear accumulation of cyclin D1 maintains a constant proliferative stress that sensitises cells to secondary genetic alterations, a process which is likely to promote carcinogenesis. The molecular mechanisms responsible for nuclear accumulation of cyclin D1 (and potentially of carcinogenesis) are under investigation, and several studies have already provided interesting data. Again, Diehl and his group had shown that GSK-3β kinase has an important function in the nuclear export of cyclin D1, and speculated that deregulation of GSK-3β may be an early event in carcinogenesis [39]. It was also shown that Frat1, a GSK3-binding protein, controls the nuclear export of GSK-3β. The deregulation of Frat1 in cancer should now perhaps be investigated [40]. Most interestingly, sequencing of the CCND1 gene in a series of endometrial cancers revealed mutations that could block cyclin D1 nuclear export [41]. Two endometrial carcinomas were found to have a single-base substitution in CCND1 that changed proline 287 to threonine and serine, which can interfere with the phosphorylation of Thr 286 by GSK-3β. In another endometrial carcinoma, a 12-bp in-frame deletion that eliminated amino acids 289–292 was detected, which may affect the binding of cyclin D1 by Crm1. More recently, sequencing of CCND1 exon 5 in oesophageal cancer revealed additional mutations which can specifically disrupt phosphorylation of cyclin D1 at Thr 286, thereby leading to nuclear accumulation of cyclin D1. The cancer-derived cyclin D1 alleles induced transformation and showed induced ability to associate with and activate CDK4 [42]. Studies to assess the prevalence of these specific mutations in lung cancers are ongoing.

Additional evidence linking cyclin D1 nuclear accumulation with carcinogenesis originated from the identification of the cyclin D1b isoform. As explained in detail in the next section, cyclin D1b is a result of alternative splicing of the primary cyclin D1 transcript, and was shown to act as an independent oncogene [43], [44]. Cyclin D1b binds to and activates CDK4, but lacks the fifth exon containing both the GSK-3β phosphorylation site Thr 286 and the Crm1 binding site, leading to the constitutive nuclear localization of cyclin D1b. Finally, it has been hypothesized that perturbation of the process of ubiquitination might also impact on the nuclear accumulation of cyclin D1. By analogy with a cyclin E ubiquitin ligase named archipelago, which is mutated and inactivated in human cancer cell lines expressing high levels of cyclin E, it is possible that defects in the degradation-process of cyclin D1 may contribute to its accumulation in tumour cells [45]. Supporting this hypothesis, Russell and colleagues showed that a defect in the SCF ubiquitin ligase complex, which associates with cyclin D1, may occur in 15–20% of breast cancers and may result in the elevation of cyclin D1 expression levels [46].

3. CCND1 polymorphism 

As discussed, the deregulation of CCND1 resulting in inappropriate gene expression can occur in human tumour cells as a consequence of gene amplification or chromosomal translocation. The discovery of such chromosomal-level events in a range of human malignancies raised the possibility that like other genes which are mutated in tumours, for example KRAS and EGFR, CCND1 might also be deregulated in a subset of lesions through point mutations conferring novel properties onto the encoded protein. In 1995, exploring this hypothesis, two groups published mutation scans of CCND1 in tumour types within which gene amplification and overexpression of cyclin D1 had previously been seen, breast carcinoma and NSCLC [32], [47]. Both groups failed to identify activating coding sequence mutations in the gene, suggesting that such events are rare or do not occur in these tumour types. However, both groups reported the identification of a previously unknown and common single nucleotide polymorphism (SNP) involving the last base of the fourth exon of the five-exon CCND1 gene. This polymorphism (A or G at nucleotide 870 of the coding sequence) did not change the specified amino acid, it did however modify a natural restriction site for the ScrFI enzyme, a property which was to prove useful in assessing the impact of this SNP on CCND1 regulation (Fig. 2).

View Large Image Download to PowerPoint

Fig. 2. Alternative splicing of the cyclin D1 transcript and functional consequence. Organization of the CCND1 gene and alternative splicing into transcript a and b is shown (leader sequence is depicted in black). Splicing is modulated by a frequent CCND1 gene polymorphism: presence of the G870 or the A870 equally leads to splicing into transcript a, which consists of exons 1–5 and encodes cyclin D1(a). Exon 5 comprises Thr 286, a site which is required for cyclin D1 nuclear export by Crm1. Presence of the A870-allele favours splicing into transcript b that encodes for cyclin D1b. Cyclin D1b, which lacks exon 5-encoded Thr 286, has been shown to accumulate in the nucleus, and to induce malignant transformation in normal cells. Abbreviations: Thr, threonine.

Betticher et al. hypothesised that while the A870G SNP did not alter the coding sequence, its position adjacent to the splice site, within the splice donor consensus sequence, might result in some difference in the efficacy of splicing of the primary CCND1 transcript [47]. Specifically, they explored whether the presence of an A at nucleotide 870 might reduce the efficiency of splice-site recognition in the unprocessed mRNA and result in a failure to correctly splice exons 4–5. This hypothesis was tested using an RFLP (restriction fragment length polymorphism) reverse transcription-polymerase chain reaction (RT-PCR) procedure. A reverse RT-PCR primer was designed to intron 4 sequence immediately downstream of exon 4. Using forward primers in a number of upstream exons, they were able to show by conventional and RACE (rapid identification of cDNA ends) RT-PCR that a novel, polyadenylated CCND1 transcript (transcript b), which included exons 1–4 and intron 4 sequence downstream of exon 4 was expressed in all tissues and cell lines examined, where the conventionally spliced cyclin D1 transcript could be detected. Somewhat surprisingly, this transcript was present regardless of the specific A870G genotype. In the first instance, these analyses therefore confirmed that CCND1 was alternatively spliced in at least some normal human tissues.

Although originally thought to be a relatively uncommon phenomenon, we now know that the majority of human genes show some degree of alternative splicing, [48] with this splice variability adding product complexity to the output from the unexpectedly small number of human genes. Although the detection of alternative splicing of CCND1 might therefore have been of no particular significance in the context of human cancer, two further observations suggested that there was some clinical relevance associated with this phenomenon; firstly, the A870G polymorphism appeared to modulate the frequency with which the exons 4–5 splice failed to occur in normal lung and tumour tissue and secondly, one allele, A870, was associated with a significantly worse relapse-free survival time in NSCLC patients [47].

Transcript-specific, allele-specific expression analysis in A870G heterozygous lung cancer patient tissues by RFLP-RT-PCR demonstrated that there was a difference in the relative balance of transcripts a and b encoded by the A870 compared with the G870 allele. Considering first the conventionally spliced transcript a, approximately equal levels of this splice variant were encoded from the A and the G alleles in a range of normal lung and tumour tissues. However, when the allelic origin of transcript b was determined in those same samples, the RFLP-RT-PCR data suggested that in each case, the variant mRNA was predominantly spliced from the A allele. These data imply that transcript b is at a lower level relative to transcript a in these tissues, given that transcript a is spliced at approximately equal levels from the A and G alleles. Such a conclusion is supported by our own quantitative transcript-specific RT-PCR analyses in NSCLC (previously unpublished results, Fig. 3). Holley and colleagues were able to confirm the tendency for transcript a to be encoded equally from both the A870 and G870 alleles and transcript b to be spliced predominantly from the A870 allele in a second malignancy, squamous cell carcinoma of the head and neck and corresponding normal tissue [49]. Interestingly, using a competitive multiplex RT-PCR analysis to investigate the balance of transcript a versus b in mantle cell lymphoma, chronic lymphocytic leukaemia and normal leukocyte samples, Howe and Lynas demonstrated that transcript b was the predominant splice variant in 22/60 (37%) cases [50]. Relative transcript balance appeared to be unrelated to genotype although transcript-specific, allele-specific expression analyses were not reported. This latter study suggests that while under certain conditions, in certain tissues transcript b is a minor product encoded predominantly from the A870 allele, in other tissues the splicing pattern of the gene may be substantially altered such that the variant transcript is the major product of the gene, presumably from either the A870 or G870 allele. This further implies that the probability of the exons 4–5 splice of CCND1 may be modified within specific cell types or situations by the involvement of unidentified splicing factors.

View Large Image Download to PowerPoint

Fig. 3. Cyclin D1 transcript levels in NSCLC. RNA was extracted from 48 NSCLC samples, and quantitative real time RT-PCR was performed using specific primers for cyclin D1 transcript a, cyclin D1 transcript b, and 7S RNA (details available upon request). Mean values of three measurements are shown. Cyclin D1 transcript levels were normalized to 7S RNA, and tumours were ranked according to their relative transcript a level. Transcript b was detected in all NSCLC samples, although at lower levels (median factor of 21.7) than transcript a. No correlation between cyclin D1 transcript levels and A870G genotype (not shown) was found in these patients. Abbreviations: PCR, polymerase chain reaction; RT, reverse transcription.

These early studies demonstrated that CCND1 was alternatively spliced, that the level of alternative splicing was modulated to some degree by the A870G polymorphism (or another polymorphism in linkage disequilibrium with this SNP) and that there was some evidence that the A870 allele was associated with a poor clinical outcome in NSCLC patients. Transcript b is predicted to encode a variant cyclin D1 protein (cyclin D1b) with an altered carboxy-terminus. Indeed, a number of studies have shown that this protein product has different biological properties when compared with the conventional cyclin D1a protein. Lu and colleagues and Solomon and colleagues demonstrated that in contrast to the conventional cyclin D1a protein, cyclin D1b is constitutively nuclear, presumably as it lacks exon 5-encoded Thr 286, the residue critical for nuclear export [43], [44]. Solomon and colleagues also showed that cyclin D1b was able to induce transformation in NIH 3T3 cells, in contrast to cyclinD1a. Holley and colleagues further demonstrated that cyclin D1 knockout mouse fibroblasts demonstrated a significantly enhanced ability to form colonies in soft agar growth when transfected with cyclinD1b-expressing compared with cyclinD1a-expressing constructs [51].

Although these experimental model-based studies are highly suggestive, the principle way in which the functional significance of a polymorphic variant can be confirmed is through molecular epidemiological analysis. If genotype is consistently correlated with either incidence, the clinical course of disease or the response to treatment, then this provides strong evidence that the variation associated with that polymorphism is of biological and perhaps clinical importance. The early functional and epidemiological data suggested that A870G genotype might modify disease outcome in lung cancer. This initial observation, coupled with the key importance attributed to the deregulation of CCND1 in human cancer, prompted a series of investigations examining the association of this SNP with disease risk and/or outcome in a range of different malignancies. One problem concerning the evaluation of reported data relating to whether a SNP associates with disease parameters rests in the problem of publication bias. This manifests as a tendency for journals to be less willing to publish negative data. It is therefore generally the case that only after a large number of often relatively small association studies or else a smaller number of large studies have been published that a pattern can be seen. In the case of the CCND1 A870G polymorphism, a consistent pattern has now emerged in that the A870 allele tends to be associated either with an increased risk of disease, an increase in severity, an earlier age of onset or an increased risk of developing premalignant lesions [52]. In their recent review, Knudsen and colleagues described 43 separate studies which examined associations between the CCND1 A870G genotype and cancer risk or outcome. Overall, the majority of the data were consistent with the A870 genotype being associated with increased risk and poor outcome [53]. Whilst these data strongly support the argument that the A870G SNP may modify disease risk or outcome, the additional risk in these studies associated with genotype tended to be relatively small and therefore the clinical utility of this SNP considered in isolation is not likely to be particularly significant. The relevance of A870G to lung cancer is supported by the results of several case-control studies (Table 1). In the Chinese population, Qiuling and colleagues conducted a study of 182 cases and 185 controls, and reported that the AA genotype was associated with moderately increased risk of lung cancer (OR 1.87, 95% CI 1.01–3.45) [54]. In the US population, the study by Buch and colleagues included 273 upper aero-digestive tract cancer cases (109 with lung cancer) and 269 controls, and showed elevated risk for AA/AG (OR 2.1, 95% CI 1.5–2.6) [55]. Another US study, including 781 cases and 781 controls, was recently presented by Spitz and colleagues and resulted in moderately elevated risk in the presence of the AA genotype (OR 1.35, 95% CI 1.1–1.7) [56]. A fourth study in Europe by Gautschi and colleagues included 244 cases and 187 controls, and did not detect significant differences in genotypes between cases and controls [57].

Table 1. CCND1 A870G gene polymorphism studies in lung cancer

	Reference	Study	N	Study population	Study aim	Result	P/OR (95% CI)
	[33]	Betticher (1995)	68	Cases only	Outcome	Poor progression free survival for AA/AG	P=0.04
	[54]	Qiuling (2003)	367	182 cases and 185 controls	Risk	Elevated risk for AA	OR=1.87 (1.01–3.45)
	[55]	Buch (2005)	542	273 upper aero-digestive tract cancer cases (including 109 lung cancers) and 269 controls	Risk	Elevated risk for AA/AG	OR=2.1 (1.5–2.6)
	[56]	Spitz (2006)	1562	781 cases and 781 controls	Risk	Elevated risk for AA	OR=1.35 (1.1–1.7)
	[57]	Gautschi (2006)	431	244 cases and 187 controls	Risk	No genotype difference between cases and controls, but association between genotype and smoking among cases	P=0.007
					Outcome	Poor chemotherapy response rate for AA/AG	P=0.04

Abbreviation: OR, odds ratio.

In summary, we can confirm that cyclin D1 is alternatively spliced. We can also confirm that the frequency of alternative splicing can be modulated by an intragenic DNA sequence polymorphism and that this polymorphism is to some extent predictive of the risk of malignant disease, clinical outcome or response to treatment for a range of tumour types. However, as is generally the case for common genetic variation, such associations tend to be weak and their clinical utility is so far not proven or even strongly implied [58]. To date, there are no convincing data which suggest that the specific targeting of the variant cyclin D1b protein would significantly reduce the risk of future disease or improve the treatment of invasive malignancy.

4. Deregulation of cyclin D1 in NSCLC 

There are different levels of evidence that implicate a particular gene product in some aspect of the malignant phenotype. Paradoxically, while epidemiological data may suggest that constitutional variation in a candidate gene modifies disease risk, the associated gene product may not be especially useful in terms of targeting anti-cancer therapeutics. This is particularly true for sequences contributing to disease predisposition through loss of function (the tumour suppressors). Replacement of missing function is pharmacologically difficult and may in some cases be therapeutically irrelevant in the context of killing fully developed tumour cells.

Epidemiology aside, perhaps the strongest evidence for causality or involvement in carcinogenesis of a candidate sequence is the occurrence and clonal selection of somatic mutations in that particular gene, especially where those mutations demonstrably positively alter the function of the encoded protein (for example, as seen with KRAS or EGFR). If such mutations are not detected in human tumours, as may be the case for CCND1 in NSCLC, indirect genomic evidence associated with a loss of the control of gene expression can still suggest a role in the cancer process. However there is a level of uncertainty in such data, largely due to the fact that genomic events such as chromosome gains, gene amplifications and chromosomal translocations potentially, in any one instance, involve changes in the regulation or expression of a variable number of different genes. Adding complexity to this issue, with respect to whether a particular level of observed expression in a tumour is appropriate or pathological, we generally know very little about the identity or gene expression pattern of the tumour cell progenitors. While such concerns are important when we are attempting to identify the key pathological changes associated with the switch from normal to neoplastic behaviour, they are less of a concern if our primary objective is to identify prognostic or predictive markers.

Data implicating cyclin D1 in lung cancer pathogenesis fall into two types (Table 2A, Table 2B, Table 2C). Firstly, a number of studies in NSCLC have shown that the CCND1 locus at 11q13 is amplified in up to 32% of cancers [33], [59], [60], [61] and secondly, immunohistochemical (IHC) analysis has shown that the cyclin D1 protein is present at high levels in 12–76% of invasive cancer and in a significant fraction of non-invasive lesions of the bronchial epithelia of lung cancer patients [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87]. Gene amplification is a mutational change which may be present in a tumour cell as a consequence either of chance (essentially random, non-selected mutation occurring historically in the progenitor) or else because the amplification event confers a growth advantage onto the host cell which consequently, over time, undergoes clonal selection. The relatively frequent detection of CCND1 amplification in NSCLC, in separate studies, argues against the chance explanation. However, although it would be expected, it is not clear from such data taken in isolation that CCND1 is the sequence responsible for conferring the presumed selectable growth advantage. Indeed in breast cancer, mapping of 11q13 amplicons has suggested the existence of at least four cancer-relevant targets in the region [88].

Table 2A. CCND1 gene copy number and cyclin D1 expression studies in lung cancer

	Reference	Study	N	Increased gene copy number N (%)	IHC positive N (%)	Associations	P-value
	[33]	Betticher (1996)	53	8 (15)	25 (47)	Expression by IHC correlated with
						Reduction in local relapse rate	0.01
						Less lymphocytic infiltration of the tumor	0.02
						Poorly differentiated histology	0.04
	[59]	Marchetti (1998)	57	18 (32)	25 (44)	Increased gene copy number correlated with expression	0.0046
						Increased gene copy number was associated with poor survival	<0.01
	[60]	Mishina (1999)	111	0 (29 analyzed)	13 (12)	Expression was associated with longer survival	0.045
	[61]	Reissmann (1999)	298	14 (5%)	No IHC	Increased CCND1 gene copy number was associated with increased EGFR gene copy number	<0.005

Table 2B. Cyclin D1 expression studies in lung cancer

	Reference	Study	N	IHC positive N (%)	Associations	P-value
	[62]	Yang (1996)	102	18 (18%)	No significant association with survival
	[63]	Kwa (1996)	96	55 (57%)	No significant association with survival
	[64]	Mate (1996)	56	24 (43%)	Association with poor differentation	0.01
	[65]	Caputi (1997)	60	30 (50%)	Overexpression correlated with survival	0.003
	[66]	Nishio (1997)	208	81 (39%)	Absence was associated with poor survival	0.01
	[67]	Lingfei (1998)	104	72 (70%)	Co-expression with cdk4
	[68]	Tanaka (1998)	101	45 (45%)	Co-expression with Ki-67
	[69]	Brambilla (1999)	168	73 (43%)	Rb-negative/p16-positive/cyclin D1-negative was associated with poor survival	0.002
	[70]	Volm (1999)	181	130 (72%)	Expression was higher in smokers than in non-smokers	<0.01
	[71]	Malusecka (1999)	112	34 (30%)	Expression was more frequent in stage I than in stage III	0.03
	[72]	Keum (1999)	69	24 (35%)	Expression was associated with advanced stage and poor survival	0.006
	[73]	Anton (2000)	467	(graded)	Absence was associated with poor prognosis in squamous carcinoma patients	0.025
	[74]	Nguyen (2000)	89	45 (51%)	No significant association with clinicopathological parameters
	[75]	Gugger (2001)	92	40 (43%)	Increased risk of relapse for cyclin D1 negative tumors	0.02
	[76]	Yamanouchi (2001)	157	58 (37%)	No significant association with histology or grade
	[77]	Jin (2001)	106	48 (46%)	Expression was associated with poor survival	0.0002
	[78]	Ikehara (2003)	72	26 (36%)	Expression was associated with poor survival	0.0049
	[79]	Esposito (2005)	105	63 (60%)	Expression was associated with poor survival	<0.0001
	[80]	Dworakowska (2005)	111	55 (49%)	Cyclin D1-negative/p53-positive was associated with poor survival	0.027
	[81]	Wang (2006)	68	52 (76%)	Expression correlated with lymph node metastasis and histology of squamous cell carcinoma	<0.05

Table 2C. Cyclin D1 expression studies in pre-invasive lung lesions

	Reference	Study	N (pts)	IHC positive sites (%)	Histology	Associations	P-value
	[82]	Betticher (1997)	33	25 (76%)	Ten normal epithelia, 19 hyperplasias and four carcinomas in situ
	[83]	Kurasono (1998)	34	7/15 (47%)	Low grade atypical hyperplasia	Expression was lower in carcinomas than in preinvasive lesions
				8/9 (89%)	High grade atypical hyperplasia
				3/11 (28%)	Early adenocarcinoma
				8/23 (35%)	Overt adenocarcinoma
	[84]	Brambilla (1999)	75 (+22 controls)	3/45 (7%)	Hyperplasia	Expression correlated with histological grade	P=0.0001
				2/30 (7%)	Metaplasia
				2/23 (9%)	Mild dysplasia
				12/26 (46%)	Moderate dysplasia
				15/41 (37%)	Carcinoma in situ
	[85]	Lonardo (1999)	Not reported	0/36 (0%)	Normal epithelium	Expression was detected in preinvasive lesions before the development of invasive carcinoma
				2/28 (7%)	Metaplasia
				5/34 (15%)	Epithelial atypia
				3/17 (18%)	Low-grade dysplasia
				14/30 (47%)	High-grade dysplasia
				15/36 (42%)	Squamous cell carcinoma
	[86]	Ratschiller (2003)	48	167/288 (58%)	Normal epithelium, hyperplasia, dysplasia and carcinoma in situ	Nuclear expression increased with histological grade	P<0.0001
				22/48 (46%)	Invasive cancer
	[87]	Lantuejoul (2005)	27	2/21 (9%)	Normal epithelium or hyperplasia	Expression correlated with histological grade	P=0.002
				2/17 (11%)	Squamous metaplasia
				4/18 (22%)	Mild dysplasia
				5/16 (31%)	Moderate dysplasia
				6/17 (35%)	Severe dysplasia
				11/33 (33%)	Carcinoma in situ
				14/27 (51%)	Invasive cancer

Abbreviations: IHC, immunohistochemistry; pts, patients.

Strengthening the argument that CCND1 contributes to lung carcinogenesis however are IHC data suggesting that cyclin D1 expression is variable in tumours and that clinical outcome may relate in some way to the expression pattern. Such data are often difficult to compare across studies as different antibodies may be used and different criteria may be employed to determine whether a particular level of expression is appropriate or abnormally high (or low). This may explain the wide range (12–76%) described in these analyses for cyclin D1 positivity, with some studies scoring expression as positive or negative, and others scoring overexpression against an implied or assumed normal level [33], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80]. In all studies, it is clear that a fraction of NSCLCs express high levels of CCND1. Tumours with clear CCND1 gene amplification tend to fall mostly into this class, adding weight to the hypothesis that such expression is aberrant and pathological [33], [59], [61]. In some NSCLC studies, high levels of cyclin D1 were associated with a more favourable clinical outcome, in others a worse outcome and in some, no associations have been detected. In contrast to breast cancer, expression of the cyclin D1 protein was reported to be predominantly cytoplasmic in a number of studies. The reason for this is not clear and may reflect changes to other proteins associated with cyclin D1 function, such as the CDK inhibitor p21 [89]. Taking these data together, there is therefore a strong rational justification for large, definitive, internally consistent IHC studies to further investigate associations of cyclin D1/CCND1 expression and clinical outcome. Such analyses, if accounting for additional confounding factors, including the presence or absence of functional Rb protein, would help to clarify the role of deregulation of cyclin D1/CCND1 expression in malignancy.

The observation that CCND1 was amplified and overexpressed in a fraction of tumours and strongly expressed when compared with normal lung in a larger fraction of carcinomas without gene amplification was strong evidence implicating the inappropriate expression of cyclin D1 in lung carcinogenesis. Further evidence from the study of Betticher et al. [33] provided data consistent with this hypothesis. This analysis demonstrated that when the protein was overexpressed, there was invariably a corresponding allele-specific expression imbalance in the levels of the mRNA transcript in the tumour tissue (transcript a), an observation consistent with inappropriate, pathological upregulation (or splicing) of one allele. A number of groups subsequently examined at which grade of malignant transformation cyclin D1 might be deregulated [82], [83], [84], [85], [86], [87]. NSCLCs arise from the bronchial epithelium, which in a chronic smoker is repeatedly, over years, exposed to the carcinogenic compounds present in cigarette smoke. Tumour-free bronchial epithelia and pre-invasive lesions from lung cancer patients often expressed high levels of cyclin D1, observations which further support a role for CCND1 deregulation in bronchial neoplasia. In a more detailed analysis of this phenomenon, Ratschiller and colleagues made an extensive investigation of cyclin D1 expression, CCND1 allelic imbalance (AI) and CCND1 transcript a allelic expression imbalance (AEI) in 288 geographically-distinct bronchial epithelial sites from 48 patients [86]. The analysis demonstrated that cyclin D1 was overexpressed, mostly in the cytoplasm, in 167 (58%) of those sites. In 69 of 144 analyzed sites, in particular in those overexpressing cyclin D1 (P=0.006), there was an AEI favouring transcripts encoded from one or other allele. In 12 of 24 informative (heterozygous) patients, the presence or absence and degree of AEI was consistent at all sites examined (generally six sites per patient) and in the other 12, it was variable in one or more sites. Of the 12 patients with consistent allelic expression ratios, two were balanced at all sites, including the primary tumour and no cyclin D1 expression was detected in any sample. This likely represents a normal situation with respect to cyclin D1 expression. The other 10 patients showed a consistent AEI, favouring the same allele at all sites, including in seven patients, the primary tumor. Nuclear cyclin D1 expression in at least one site was associated with a history of heavy smoking (P=0.02) and shorter overall survival (P=0.01) (Fig. 3, Fig. 4). Ratschiller and colleagues concluded that allele-specific, damage-driven, deregulation of cyclin D1 likely preceded and perhaps facilitated the spread of pre-neoplastic clones of cells across the bronchial epithelial surface. An alternative and perhaps less likely explanation is that cyclin D1 is normally expressed predominantly from one parental allele in one or more components of the bronchial epithelium. Further experiments, especially in bronchial tissue from non-smokers would be needed to clarify the mechanism responsible for this observation. Nevertheless, these experiments provide a further level of evidence implicating the deregulation of CCND1 as a pivotal early mechanism in bronchial neoplasia.

View Large Image Download to PowerPoint

Fig. 4. Cyclin D1 immunostaining. Peroxidase–antiperoxidase technique (bar=50μm). (A) Adenocarcinoma (non-small cell lung cancer) with nuclear cyclin D1 overexpression. (B) Bronchiolar epithelium with nuclear cyclin D1 overexpression in hyperplastic basal cells (on the left) and diffuse cytoplasmic overexpression. (C) Squamous cell carcinoma (non-small cell lung cancer) with heterogeneous cytoplasmic cyclin D1 overexpression. (D) Bronchiolar epithelium with cytoplasmic cyclin D1 overexpression. (A) and (B) illustrate the typical findings in a heavy smoker and (C) and (D) in a non-smoker [86].

In summary, most of the available data on gene amplification and expression in invasive NSCLC, precursor lesions and normal bronchial epithelia from lung cancer patient studies are consistent with the hypothesis that the deregulation of CCND1 and its product, cyclin D1, is an important mechanism associated with bronchial carcinogenesis.

5. Clincial perspectives 

The insights into the molecular biology of cyclin D1 described perhaps allow new approaches to the diagnosis and treatment of lung cancer. Gautschi and colleagues reported on the CCND1 A870G polymorphism and its potential use for the identification of smokers who are at high risk for lung cancer development, as well as for the selection of NSCLC patients for platinum-based chemotherapy [57]. The study included 244 cases, with the majority (81%) of these patients having stages III–IV disease. Genotyping, performed using patient blood and RFLP-PCR, revealed two new findings. Firstly, lung cancer patients with the GG genotype had a history of a significantly higher amount of cigarette smoking at the time of diagnosis than patients with AA/AG genotype, which suggested that individuals with AA/AG genotype may be more susceptible to tobacco-related carcinogens than individuals with the GG genotype. Secondly, the disease control rate following platinum-based chemotherapy was significantly higher in patients with GG genotype than in patients with AA/AG genotype. Although it would be premature to make any recommendations regarding the clinical use of the CCND1 A870G polymorphism at the present time, the data are consistent with the current evidence, indicating that the CCND1 A870G polymorphism is not only biologically important, but may also be also clinically relevant. It is tempting to speculate that in the future, A870G status might be one element in a genotypic-profiling tool that can be used to assess disease risk (allowing the rational targeting of smoking cessation resources) or to select the most appropriate anti-cancer treatment regimens.

The active cyclin D1/CDK complex is an interesting drug target, as demonstrated by cyclin D1 antisense experiments, which resulted in decreased proliferation in human oesophageal cancer cells in vitro and in vivo [90]. Due to the clinical problems with antisense approaches, much of today's drug development in this field concentrates on the identification of small-molecule kinase inhibitors, which are stable, tolerable and active in patients. The non-enzymatic cyclin D1 molecule does not fall into the class of proteins that are conventionally considered to be targetable by such small molecule inhibitors. However, its catalytic partners, the CDKs, and in particular CDK4, should be amenable to pharmacological inhibition. Because of the similarity in the ATP-binding pocket among the different CDKs, early attempts to identify such targeted compounds mostly revealed pan-CDK inhibitors, with the most prominent member of this class being flavopiridol. In a phase II trial, 20 patients with previously untreated stages IIIB–IV NSCLC were treated with a 72h continuous infusion of flavopiridol every 14 days at a dose of 50mg/m2/day [91]. The most common toxicities included grades 1 or 2 diarrhoea in 11 patients, asthenia in 10 patients, and venous thromboses in 7 patients. At 8 weeks, 18 patients were evaluable for response, 10 (55%) patients had stable disease, and 8 had progressed. No objective responses were observed, consistent with the physiological function of the CDKs as drivers of cell growth. Flavopiridol was studied in a phase I trial in combination with paclitaxel and carboplatin, with a documented 36% response rate and a 66% disease control rate [92]. Another phase I study tested the combination of flavopiridol and docetaxel, and reported a disease control rate of 63% [93]. Phase II/III trials with flavopiridol and chemotherapy are ongoing, and other pan-CDK inhibitors with expected lower toxicity are on the way.

Small molecule inhibitors that are highly specific for CDK4/6 have recently been reported. PD-0332991 is a pyridopyrimidine with high selectivity for D cyclin/CDK complexes, including cyclinD1/CDK4 (isolated enzyme IC50=11nmol/L), cyclin D3/CDK4 (IC50=9nmol/L), and cyclin D2/CDK6 (IC50=15nmol/L) [94]. PD-0332991 induced G1 arrest and reduced Rb protein-phosphorylation in cancer cells, and as expected, this effect was restricted to cells which express Rb (for example, marked sensitivity of the Rb-positive H1299 lung cancer cell line, and relative resistance of the Rb-negative H2009 lung cancer cell line). Oral administration of PD-0332991 at 150mg/kg daily for 14 days to mice bearing human tumour xenografts (colon, breast, prostate and lung cancer) resulted in tumor regression, inhibition of Rb-phosphorylation and inhibition of transcription of genes controlled by E2F, such as CCNE2 (cyclin E2) and TOP2A (topoisomerase IIα). A phase I study to test the safety of PD-0332991 and to evaluate two oral schedules in patients with advanced cancer is ongoing, and it will be important to address the question whether patients should be selected for treatment based on Rb expression in tumours. Other selective cyclin D1/CDK4 inhibitors have also been presented as lead-compounds for further clinical drug development [95]. Finally, a broad number of other compounds targeting mitogenic signalling pathways can indirectly attenuate cyclin D1 expression to mediate cell cycle arrest.

6. Conclusion 

Human lung carcinogenesis is a complex mechanism at the molecular level, with many different oncogenes (e.g. CCND1, KRAS, EGFR, MYC, STK15) and tumour-suppressor genes (RB1, TP53, CDKN2A, FHIT, RASSF1) involved. It has been demonstrated that smoking induces nuclear accumulation of cyclin D1 in human bronchial epithelium in situ, and that nuclear accumulation of cyclin D1 induces uncontrolled proliferation in normal human cells, which may facilitate the development of invasive cancer. The recent identification of mutations associated with cyclin D1 nuclear accumulation in oesophageal cancer warrant a new and more intensive mutation scan in lung cancer. Novel approaches to the diagnosis and treatment of lung cancer appear to emerge from the insights into the biology of cyclin D1, including use of the CCND1 A870G polymorphism as a biomarker, and the development of small-molecule kinase inhibitors selectively targeting the cyclin D1/CDK4 complex.